An Integrated Design Method for Elderly-Friendly Game Products Based on Online Review Mining and the BTM–AHP–AD–TOPSIS Framework

Abstract

With the increase in the global aging population, the demand for elderly-friendly game products is growing rapidly. To address existing limitations, particularly in user demand extraction and design parameter setting, this study proposed a design framework integrating the BTM–AHP–AD–TOPSIS methods. The goal was to accurately identify the core needs of elderly users and translate them into effective design solutions. User reviews of elderly-friendly game products were collected from e-commerce platforms using Python 3.8-based web scraping. The Biterm Topic Model (BTM) was employed to extract user needs from review texts. These needs were prioritized using the Analytic Hierarchy Process (AHP) and translated into specific design parameters through Axiomatic Design (AD). Finally, the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) was applied to comprehensively evaluate multiple design schemes and select the optimal solution. The results demonstrate that the proposed design path offers a holistic method for progressing from need extraction to design evaluation. It effectively overcomes previous limitations, including inefficient need extraction, limited scope, unclear need weighting, and unreasonable design parameters. This method enhances user acceptance and satisfaction while establishing rigorous design processes and scientific evaluation standards, making it well suited for developing elderly-friendly products.

1. Introduction

As the global trend of population aging accelerates, improving the quality of life of the elderly has become a critical social issue. According to data released by the National Bureau of Statistics of China, the population aged 60 and above reached 310.31 million by the end of 2024, exceeding 300 million for the first time, and it is projected to rise to 32.9% by 2050, signaling the transition to a super-aged society. The decline in physical and mental functions among older adults has led to a growing reliance on health management, caregiving support, and spiritual or recreational products, promoting the rapid development of aging-friendly solutions. Driven by the “active aging” policy direction, elderly users are increasingly engaging with cultural and entertainment products. A survey conducted by the China Consumers Association indicated a notable increase in the acceptance of gaming products among the younger elderly population. Similarly, Alibaba’s Elderly Digital Life Report highlights that the compound annual growth rate of online consumption by silver-haired users has reached 20.9% over three years, with entertainment products emerging as a particularly popular category. As a significant component of spiritual and cultural consumption, game products are witnessing rising demand among elderly users. However, most existing games are designed primarily for younger users in terms of functionality and interface interaction [1,2]. This misalignment has resulted in a poor user experience, insufficient adaptability for older adults, and a high degree of product homogeneity in the aging-friendly gaming market.

Current research on the design of gaming products for the elderly primarily focuses on functional development and technological enhancements or on utilizing games as tools for cognitive training and physical rehabilitation. In terms of functionality and technological applications, Dong [3] explored psychological intervention pathways for older adults and examined the effects of gamified VR exercise on life expectancy, concluding that such game products could potentially extend the lifespan of elderly users. Chen et al. [4] developed a chess player recognition and automatic placement system, creating a Chinese chess robot by advancing image recognition and other hardware-related technologies. Ghorbani et al. [5] designed and evaluated an intelligent assistive system that integrates AR with serious games, aiming to provide multilayered support for older adults with mild cognitive impairment and their caregivers. Although these studies have made notable progress in technical applications and functional validation, they lack a scientific mapping mechanism that links the real needs of elderly users to product functions, highlighting the need for more systematic design methodologies. In the domain of cognitive training and physical rehabilitation, Alaa Abd-alrazaq et al. [6] proposed design and evaluation strategies for serious game-based interventions to enhance attention in cognitively impaired older adults based on systematic reviews and meta-analyses. Müller et al. [7] conducted experimental research on motion-sensing games from the perspective of how different game types and task difficulties affect brain and physical responses, suggesting that task design should consider its impact on brain and physical activity in elderly training. Morán et al. [8] applied a context-embedded design methodology to develop cognitive stimulation games and service strategies tailored to older adults, targeting everyday cognitive training needs. Tseng et al. [9] employed a service experience insight method to explore the cognitive needs of elderly individuals with mild to moderate dementia. The above studies demonstrate that games exhibit positive intervention effects in enhancing cognitive abilities and stimulating physical function in older adults, highlighting their high feasibility and application value. However, current research lacks game evaluation tools and design techniques specifically tailored to the elderly population. Greater emphasis should be placed on user-centered game design approaches, as existing studies remain insufficient to thoroughly explore the needs of elderly users and develop corresponding design strategies [10].

In developing game products for the elderly, accurately identifying and understanding user needs is essential for systematically optimizing functionality and user experience, thereby promoting successful market adoption. Common methods for uncovering the needs of older users include field observations, user interviews, focus groups, and questionnaire surveys, which gather behavioral data and self-reported experiences. These are typically followed by grounded theory or similar approaches to extract and synthesize product requirements. Many researchers have used semi-structured interviews and questionnaire surveys to explore user experiences and preferences among older adults in the context of game design [11,12,13]. However, these methods often suffer from low efficiency and limited scope, making it difficult to capture the full range of genuine user needs of elderly users [14]. For example, traditional interviews may be influenced by leading questions or misunderstandings, resulting in socially desirable responses that do not reflect true user intentions. As cognitive and expressive abilities decline with age, older adults may choose neutral or noncommittal responses such as “average” or “no problem” on questionnaires to avoid mental strain, leading to inaccurate representations of their actual experiences and core needs. These behavioral biases compromise data validity and reduce the effectiveness of conventional methods in extracting user needs for elderly-focused game products. As research on game design for the aging population is still in its early stages, there is an urgent need for innovative approaches that enable deeper design exploration [15].

In recent years, the rise of big data and artificial intelligence has brought significant advancements to user-demand mining and design parameter research, enabling the integration of more scientific and systematic methods. The academic community has increasingly explored data mining techniques based on online reviews to support product improvement and better understand user behavior. In the field of game design, some studies have used word frequency analysis and TF-IDF to extract keywords from reviews [16,17], whereas others have applied sentiment analysis to identify user satisfaction levels and emotional tendencies. Topic modeling methods such as Latent Dirichlet Allocation (LDA) and Biterm Topic Model (BTM) have also been employed to uncover underlying themes in textual data [18,19,20]. However, these approaches often remain limited to data collection and problem identification, lacking subsequent design solution development and evaluation [21,22]. Common challenges include unclear prioritization of user needs, vague definitions of functional parameters, and the absence of objective criteria for design evaluation. These issues can lead to arbitrary design decisions, functional mismatches, and poor user experiences, ultimately weakening product competitiveness and application effectiveness while limiting improvements in the quality of life of the aging population. Therefore, when developing game products for older adults, it is crucial to effectively mine real user needs from online reviews and ensure the scientific alignment of design parameters with a robust evaluation of the proposed solutions. To address this gap, this study proposes a design method for elderly-oriented game products based on online user reviews, establishing a systematic pathway from need identification to design generation and optimization, with the goal of enhancing both scientific rigor and user relevance to design outcomes.

This study proposed using online review data from e-commerce platforms as a key information source for user demand extraction and adopted a research approach that combines short-text topic modeling with structured design methodologies. Python-based web scraping techniques were employed to collect a large volume of user reviews across diverse brands and product types. Ensuring data diversity helps capture a wide range of user opinions and enhances the accuracy of product or service improvements [23]. The collected online reviews were subsequently analyzed and categorized. Topic modeling is widely used in text mining, with LDA and BTM being the two most commonly adopted models. Compared with traditional models such as LDA, BTM performs better on short texts such as user reviews, as it captures word co-occurrence patterns without relying on complete contextual information [24,25]. By identifying and categorizing latent topics within the text, BTM enables efficient text classification and reveals users’ primary concerns, thereby uncovering a broad range of potential user needs [26]. In this study, BTM was applied to the collected reviews to extract latent topics from which preliminary user needs were inferred. However, further screening is required to determine the relative importance of these needs. As a decision-making tool, the Analytic Hierarchy Process (AHP) was applied to quantify expert judgments into weight values [27], effectively identify key user needs, and enhance the scientific rigor and accuracy of the decision-making process. Thus, this study used AHP to extract the core requirements and key design elements of aging-friendly game products. After identifying key user needs, it is essential to translate them into specific design parameters for age-friendly game products. To accomplish this, this study integrated Axiomatic Design (AD) theory by applying an independence axiom to perform precise information mapping. This process translates user needs into design parameters using mapping matrices to systematically derive rational design parameters [28], thereby overcoming the limitations of traditional intuition-driven design processes. Finally, the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) method was used to comprehensively evaluate and optimize multiple design schemes [29]. This ensured that the final design met the core user needs and featured a well-balanced functional configuration, thereby validating the rationality of the proposed design methodology. Although existing methods have been applied to aging-related design research, they generally lack a systematically integrated pathway oriented toward practical design implementation. As a result, it remains challenging to achieve a closed-loop process that progresses from “perceiving user needs” to “functional transformation” and ultimately to “solution optimization”. To address this gap, this study proposed a BTM–AHP–AD–TOPSIS design framework based on online review texts. By integrating the strengths of BTM in short-text mining, AHP in hierarchical judgment, AD in functional mapping, and TOPSIS in optimal solution evaluation, a closed-loop design process was established, encompassing five key stages: review data collection, need identification, weight determination, design parameter mapping, and design evaluation. This study systematically extracted user needs for aging-friendly game products, mapped design parameters based on prioritized needs, and conducted corresponding solution evaluations. By grounding the design process in real user needs, the proposed method enhances the scientific rigor and practical relevance of aging-friendly game product design. It aims to improve alignment between design outcomes and user expectations while offering both theoretical insights and practical guidance for innovation in this field.

2. Methodology

This study proposes a novel design method for elderly-friendly game products, as illustrated in Figure 1, comprising four main components.

• User review data mining and topic modeling: Python-based web crawler tools were used to collect game product review data related to elderly users from e-commerce platforms to obtain real user feedback. The review texts were pre-processed through cleaning, word segmentation, and deduplication to build a reliable corpus for subsequent analysis. Subsequently, the BTM topic model was applied to extract latent topics from the review data.
• User demand identification and weight calculation: AHP was used to evaluate the relative importance of user needs identified through BTM. By constructing a two-level judgment matrix, experts were invited to conduct pairwise comparisons of each demand item to calculate weights and identify key user concerns.
• Design parameter mapping: After clarifying the key requirements, AD theory was introduced to map user needs to technical design parameters, thereby generating actionable design specifications.
• Design evaluation and verification: The TOPSIS method was used to evaluate the scheme using elderly game product design as a case study. Each scheme was ranked based on its closeness to the ideal solution. The optimal scheme was identified, thereby verifying the validity and effectiveness of the proposed design approach.

2.1. Experimental Data Sources

The experimental data were sourced from China’s JD.com e-Commerce platform. As one of the leading e-commerce platforms in China, JD.com provides a large volume of authentic user reviews, offering high representativeness and reliability, making it a suitable data source for user need extraction. This ensured the representativeness and timeliness of the samples. All review information used in this study is publicly available. An automated web crawler developed in Python utilizing the Selenium library to simulate browsing behavior and capture content was used for data collection. The extracted review data were exported and stored in Microsoft Excel for further analyses.

2.2. Participants

This study invited 17 experts related to aging-friendly products and game design to form a focus group to participate in the research work, comprising 4 experts in elderly-friendly product design, 4 PhD students in industrial design, 5 master’s students specializing in elderly-friendly product design, and 4 designers of puzzle products. Participant demographics are summarized in

Table 1. Demographic characteristics of participants.

Category	Content	Number of People	Percentage
Gender	Male	10	59%
	Female	7	41%
Age	Aged 20–25	3	18%
	Aged 26–30	5	29%
	Aged 31–40	6	35%
	Aged ≥ 41	3	18%
Education	Bachelor degree	2	12%
	Master’s degree Master’s student	8	47%
	Doctoral degree PhD student	7	41%
Occupation	Student	8	47%
	Designer	5	29%
	Academic staff	4	24%

. The focus group provided professional judgment and data support throughout the key stages of this study, including online review text analysis, feature word induction, user demand analysis, and AHP-based demand weighting.

2.3. Experimental Procedure

2.3.1. Online Review Text Mining and Topic Identification

With the rapid growth of e-commerce, analyzing online product reviews offers valuable insights into users’ genuine attitudes and helps identify the specific needs to inform design efforts. This study employed Python tools to collect and analyze user reviews, focusing on the Chinese e-commerce platform JD.com for text mining. The process consisted of review collection, preprocessing, and topic modeling. First, target products were identified, and online reviews were crawled to construct an online review database. To ensure the validity of the results, the data were preprocessed. Subsequently, the re.sub() function in Python was adopted to eliminate non-Chinese characters, followed by text processing using the Jieba library for word segmentation. Additional tasks, such as stop word removal and spoken-word normalization, were performed to prepare the data for BTM topic analysis. The detailed process is illustrated in Figure 2.

The structure and modeling process of the BTM are shown in Figure 3. This diagram clearly depicts the BTM workflow and relationships among the variables. The BTM model treats each co-occurring word pair in the text as the basic modeling unit and identifies latent topics by analyzing the co-occurrence frequency and distribution of these word pairs within the corpus. On the left side of the figure is the generative probabilistic graphical model of BTM. Here, α and β represent the Dirichlet prior distributions for the learning topic distributions. Parameters θ and $\phi$ denote the probabilities of latent topics and the distribution of word pairs for each topic, respectively. Variable k indicates the number of topics, Z represents the specific topic, W_i and W_j are individual words within a word pair, and |B| denotes the total number of word pairs in the corpus. The right side of the figure outlines the model-training process, which includes extracting word pairs from the raw text to construct the corpus, training the model, estimating parameters via Gibbs sampling, and ultimately obtaining the topic and word distributions.

To determine the optimal number of topics K, perplexity and consistency metrics were employed for evaluation. Perplexity measures a model’s uncertainty in assigning topics to a corpus, with lower perplexity indicating better model performance and training effectiveness. By calculating the perplexity for different topic numbers, the number with the lowest perplexity was selected as the final topic count. The textual data in this study were sourced from product reviews posted by elderly users on an e-commerce platform. Such reviews are typically short and unstructured. BTM, which constructs word-pair relationships, is well suited for identifying latent topics in such short texts. The probability of generating a word w in a given document d is denoted as p(w_d). The text perplexity calculation is shown in Equation (1).

$$p\left(w_d\right)={\displaystyle {\sum }_{z=1}^K} p\left({\displaystyle \frac{w}{z}}\right)\cdot p\left(z\right)$$

(1)

where w_d denotes the word vector in corpus d; z represents the latent topic index; and p(z) is the prior probability of topic z, indicating its likelihood of occurrence across the entire corpus.

After obtaining the generation probability p(w_d) for each word in the elderly user reviews, the overall perplexity P of the corpus was calculated to evaluate the BTM model’s fit to the review data, as shown in Equation (2).

$$P=exp\left(-{\displaystyle \frac{\sum _{d=1}^M lbp\left(w_d\right)}{\sum _{d=1}^M \left(N_d\right)}}\right)$$

(2)

where M is the total number of texts in the corpus, that is, the total number of reviews of old game products; and N_d is the number of words contained in the short text document d, which represents the total word count of each review.

For feature word extraction and topic summarization, the BTM model was used to analyze the topics and extract the top-ranked words with the highest probabilities as feature words representing the content of each topic. A focus group was subsequently invited to summarize and synthesize textual content, ultimately identifying user demand topics for elderly-friendly game products across various dimensions.

2.3.2. Identification of Key Requirements

In this study, AHP was applied to calculate the weights of various user needs for a product. Initially, the influencing factors at the decision-making level were analyzed, and a hierarchical model was constructed, progressing from the goal layer to the solution layer, as illustrated in Figure 4. A panel of experts was invited to evaluate each level of user need through pairwise comparisons using a scale of 1–9. The mean scores were calculated to derive judgment matrices. Based on these matrices, the geometric mean algorithm was used to calculate the weights of user needs. Consistency checks were conducted to ensure the rationality of the judgment matrix. Based on these weights, the key user needs with the highest rankings were identified, forming the basis for determining product design elements and guiding subsequent solution designs. The detailed calculation procedure is outlined in Equations (3)–(8):

• Step 1. Obtain the product of hierarchical needs M_i:

$$M_i=\prod _{j=1}^m b_{ij}=(i=\mathrm{1,2},\dots ,m; j=\mathrm{1,2},\dots ,n)$$

(3)

where b represents the user requirement comparison matrix constructed in this study. The value in the i row and j column of the matrix is denoted as b_ij, representing the relative importance of requirement i compared with requirement j. The total number of requirement indicators is denoted as m, and M_i represents the relative product value of indicator i with respect to all others, serving as the basis for weight calculation. In this study, the values of b_ij were derived from expert pairwise comparisons of demand topics extracted using the BTM model.

• Step 2. Obtain the geometric mean of the product of hierarchical needs a_i:

$$a_i=\sqrt[m]{M_i}(i=\mathrm{1,2},\dots ,m; j=\mathrm{1,2},\dots ,n)$$

(4)

where a_i represents the overall priority of each user requirement topic aggregated from all expert judgments.

• Step 3. Obtain the relative weight, W_i:

$$W_i={\displaystyle \frac{a_i}{\sum _{i=1}^m a_i}}$$

(5)

where W_i denotes the relative weight of the i user requirement calculated using the geometric mean method. This weight reflects the importance of the requirement within the overall evaluation system.

• Step 4. In this study, experts constructed a pairwise comparison matrix based on multiple user comment topics extracted using the BTM model. To evaluate the consistency of this judgment matrix, it is necessary to calculate its maximum eigenvalue, λ_max. In AHP, if judgment matrix B and its corresponding weight vector W satisfy the relationship B⋅W = λ_max⋅W, then the ratio ${\displaystyle \frac{{BW}_i}{W_i}}$ for each element should approximate λ_max. Accordingly, matrix B is first multiplied by weight vector W to obtain a new vector BW. Each element of this vector is then divided by the corresponding component of W, and the average of these ratios is calculated to estimate λ_max. This process is expressed as follows:

$${\lambda }_{\mathrm{m}\mathrm{a}\mathrm{x}}={\displaystyle \frac{1}{n}}\sum _{i=1}^n{\displaystyle \frac{{BW}_i}{W_i}}$$

(6)

where BW_i represents the i component of the product of judgment matrix B and weight vector W, and n denotes the order of the judgment matrix, which corresponds to the number of user need topics evaluated for importance in this study.

• Step 5. A consistency check was conducted to validate the judgment matrix by calculating the consistency index (CI) and the consistency ratio (CR).

$$CI={\displaystyle \frac{{\lambda }_{\mathrm{m}\mathrm{a}\mathrm{x}}-n}{n-1}}$$

(7)

$$CR={\displaystyle \frac{CI}{RI}}$$

(8)

where n denotes the order of the judgment matrix, RI denotes the random consistency index, and CR denotes the consistency ratio. When CR ≤ 0.1, the consistency check passes. At this point, expert judgments are considered valid, providing a reliable basis for the subsequent design process.

In addition, a comprehensive weight calculation was introduced at this stage to refine the previous weight analysis results. This approach helps prevent key user requirements from being overlooked during the hierarchical decomposition process, thereby enhancing the completeness and scientific rigor of the evaluation. It also provides robust data support for the subsequent extraction of design elements and optimization of product solutions. Specifically, the weight product method was used, whereby the weight of each criterion layer was multiplied by the local weight of its corresponding sub-criterion layer to derive the comprehensive weight of each sub-criterion.

2.3.3. Design Parameter Mapping

According to the principle of information independence in AD, the key customer attributes (CAs) of elderly users were effectively converted into functional requirements (FRs) of the product, which were subsequently mapped in a zigzag pattern to generate design parameters (DPs). This process establishes a structured parameter framework within the design domain, as shown in Figure 5. It not only enhances the rigor of the design logic but also improves the operability of transforming user needs into functional requirements, thereby ensuring user-friendliness and practical applicability. This study primarily applied the Independence Axiom to ensure that each functional requirement was satisfied by a distinct and independent design parameter. Based on this principle and guided by identified user needs, multiple design solutions were proposed, illustrating a diversified approach to product development within the context of aging-friendly design using AD principles. The solution process for these design parameters is shown in Equations (9) and (10):

$$\left\{F{R}_s\right\}=B\left\{D{P}_s\right\}$$

(9)

where the set of functional requirements of the product is denoted as FRs, the set of specific design parameters is denoted as DPs, and the design parameter matrix is denoted as B. The specific expressions are as follows:

$$\left[\begin{array}{c}F{R}_1\\ F{R}_2\\ \dots \\ F{R}_n\end{array}\right]=\left[\begin{array}{cccc}{b}_{11}& b_{12}& \dots & b_{1n}\\ b_{21}& b_{22}& \dots & b_{2n}\\ \dots & \dots & b_{ij}& \dots \\ b_{n1}& b_{n2}& \dots & b_{nm}\end{array}\right]\left[\begin{array}{c}D{P}_1\\ D{P}_2\\ \dots \\ D{P}_m\end{array}\right]$$

(10)

where the relationships between the relevant parameters in the matrix are denoted as b_ij.

In the AD methodology, the design matrix is commonly used to represent the relationships between functional requirements (FRs) and design parameters (DPs). Its structure directly determines whether the Independence Axiom is satisfied, thereby influencing the feasibility and complexity of subsequent design stages. Based on the matrix structure, design matrices are typically classified into three types: uncoupled, decoupled, and coupled designs, as shown in Figure 6. This figure presents standard configurations for each matrix type and explains the symbols used. “X” indicates a strong relationship between an FR and a DP, whereas “O” denotes a weak relationship. In an uncoupled design matrix, matrix A is diagonal, with “X” appearing only on the main diagonal and all other elements as “O”. In this case, each FR is independently satisfied by a unique DP without interference, thus fully meeting the Independence Axiom. In a decoupled matrix, A is lower triangular, indicating that some DPs influence multiple FRs. However, by adjusting the implementation sequence, FRs can still be achieved independently. In contrast, a coupled matrix features irregularly distributed “X” values, where multiple DPs affect multiple FRs, indicating significant coupling and violation of the Independence Axiom. Such matrices typically require redesign to reduce complexity and improve system controllability.

2.3.4. Scheme Evaluation and Optimization

This study utilized the key user requirements and design parameters mapped by AD to develop diversified elderly-friendly game products. To scientifically determine the optimal design scheme, the TOPSIS method was employed for objective evaluation and comprehensive assessment of the design alternatives. First, the demand topics extracted from the BTM model were adopted as evaluation indicators, and a seven-point Likert scale was applied to rate the extent to which each product met the corresponding indicator requirements. To improve the objectivity of the evaluation and minimize the influence of the designers’ subjective preferences, rating data from different user groups were homogenized to construct the initial evaluation matrix. Subsequently, data normalization was performed to create a standardized matrix F. According to the standard computational steps of the TOPSIS method, the distances between each design scheme and the positive and negative ideal solutions were calculated, allowing for determination of the relative closeness and final ranking of each scheme. The scheme with the closest distance to the positive ideal solution and farthest distance from the negative ideal solution was identified as the optimal solution. The detailed TOPSIS calculation steps are shown in Equations (11)–(16).

• Step 1. The data obtained from the questionnaire were averaged to build an initial evaluation matrix, denoted as f.

• Step 2. The initial evaluation matrix was normalized to obtain the standardized matrix R_ij.

$$R_{ij}={\displaystyle \frac{f_{ij}}{\sum _{i=1}^m{f}_{ij}^2}}(i=\mathrm{1,2},\dots ,m;j=\mathrm{1,2},\dots ,n)$$

(11)

where f_ij is derived from the feedback scores based on the design dimensions extracted from user comments using the BTM topic model, representing the value of the i alternative under the j criterion in the original decision matrix, and m denotes the total number of alternatives.

• Step 3. The weighted matrix u_ij was obtained based on the calculation of the target weights of each evaluation indicator.

$$u_{ij}=W_j{R}_{ij}(i=\mathrm{1,2},\dots ,m;j=\mathrm{1,2},\dots ,n)$$

(12)

where W_j represents the weight of the j criterion, which is obtained during the AHP phase as the importance weight of user needs.

• Step 4. The positive and negative ideal solutions, A⁺ and A⁻, respectively, were determined.

$$\begin{array}{rr}& M_j^{+}=max\left\{u_{1j},u_{2j},u,\dots ,u_{nj}\right\}(j=\mathrm{1,2},\dots ,n)\\ & M_j^{-}=min\left\{u_{1j},u_{2j},u,\dots ,u_{mj}\right\}(j=\mathrm{1,2},\dots ,n)\end{array}$$

(13)

Thus,

$$\begin{array}{rr}& A^{+}=\left(M_1^{+},M_2^{+},\dots ,M_n^{+}\right)\\ & A^{-}=\left(M_1^{-},M_2^{-},\dots ,M_n^{-}\right)\end{array}$$

(14)

• Step 5. The Euclidean distance is solved to obtain the distance between each scheme and the ideal solution, serving as a basis for evaluating the overall performance of each alternative. Si⁺ and Si⁻ represent the distances to positive and negative ideal solutions, respectively.

$$\begin{array}{rr}& S_i^{+}=\sqrt{{\displaystyle \sum _{j=1}^n}{\left(u_{ij}-u_j^{+}\right)}^2}(i=\mathrm{1,2},\dots ,m;j=\mathrm{1,2},\dots ,n)\\ & S_i^{-}=\sqrt{{\displaystyle \sum _{j=1}^n} {\left(u_{ij}-u_j^{-}\right)}^2}(i=\mathrm{1,2},\dots ,m;j=\mathrm{1,2},\dots ,n)\end{array}$$

(15)

• Step 6. The relative closeness coefficient Ci is calculated for each game design alternative. A higher Ci value indicates that the alternative is closer to the ideal solution and has greater relative superiority.

$$C_i={\displaystyle \frac{S_i^{-}}{S_i^{+}+S_i^{-}}}(i=\mathrm{1,2},\dots ,m)$$

(16)

The comprehensive evaluation and decision making of product design schemes is a complex multi-criteria process that integrates both subjective judgments and objective data, with outcomes that significantly influence subsequent design strategies. Therefore, to verify the stability of the ranking results produced by the TOPSIS method in this study, a comparative analysis was conducted using the Grey Relational Analysis (GRA) method. The calculation process involved normalizing the data, computing the grey relational coefficients between each indicator and the reference sequence, and then calculating the average grey relational grade for each alternative to determine the final ranking.

2.4. Data Analysis

In this study, Python-based crawler scripts were developed to collect user review data on aging-friendly game products from the JD.com platform. All data are publicly available online reviews. Following word segmentation, denoising, and removal of irrelevant information, the cleaned data were verified for validity and organized using Excel spreadsheets. The BTM method was applied to extract potential user demand themes using the Python biterm library to determine the optimal number of topics. The AHP and TOPSIS methods were subsequently used to prioritize user needs and optimize design solutions. In the AHP phase, experts were invited to evaluate the relative importance of each demand factor. Their ratings were analyzed using SPSS 25.0 to conduct consistency checks and calculate weights. A consistency ratio (CR) ≤ 0.1 was considered acceptable. For the AD theory link, Excel spreadsheets were used to construct and visualize the design matrix (FR–DP), with annotations capturing the coupling relationships between functional requirements and design parameters. In the TOPSIS analysis, both experts and elderly users assessed multiple design schemes based on key needs identified in the AHP phase. SPSS 25.0 was used to standardize the evaluation data, and the Euclidean distance between each solution and the ideal solution was calculated to determine the final priority ranking of the design schemes. Finally, the Grey Relational Analysis (GRA) method was implemented using Matlab software 2016a.

3. Results

3.1. Text Mining and Demand Analysis of Online Reviews of Elderly Puzzle Games

This study focuses on “cognitive toys for the elderly” available on the JD.com e-commerce platform. User reviews were collected from product pages associated with keywords such as “elderly toys”, “age-friendly toys”, and “entertainment products for seniors”, covering 10 categories of related products. All reviews were originally written in Chinese by JD.com users. The data collection period spanned two years, from 8 September 2022 to 8 September 2024, to ensure representativeness of the samples. A total of 32,466 raw comments were initially obtained. To illustrate the data source, examples of original product reviews are shown in Figure 7 (screenshots from two sample products), and the selected sample data are presented in

Table 2. Sample review data of elderly-friendly puzzle game products.

Text
“This definitely deserves a review. The right hand of my mother is not very flexible due to illness. She is also bored at home and her brain is becoming decreasingly flexible. Therefore, I thought of getting her this. It is really great. Elderly people like it too. It’s good for both hand and brain exercises. Highly recommended!”
2. “Great quality, very fun. Both kids and elderly can play. Fast delivery and quick logistics.”
3. “Very delicate, and the colors are beautiful. I really like it.”
4. “A small hands-on toy, purely wooden, engages both hands and the brain. Quite interesting, very suitable for elderly people.”
5. “An excellent game machine for mental training. It is very suitable for elderly people. It is harder than I expected, training reaction speed. It exercises the eyes, hands, and brain, allowing parents to live a healthy and enjoyable life.”

. A multistep data preprocessing procedure was conducted to ensure the accuracy and validity of the analysis. First, low-quality entries such as duplicate reviews and default system-generated comments were removed. Only “substantive comments” were retained, excluding short or vague responses (e.g., “good” and “not bad”) consisting of fewer than four Chinese characters. After this initial cleaning, 30,738 valid comments remained in the usable corpus data. Subsequently, the re.sub() function in Python was employed to remove all non-Chinese characters from the text.

Jieba word segmentation was used to optimize the structure of the textual data to address the colloquial and diverse nature of online reviews on e-commerce platforms. Given the frequent use of non-standard expressions in aging-related reviews, such as elderly-specific slang and colloquial phrases like “练脑力” (mental exercise), “解闷” (to kill boredom), and “琢磨” (to think over), the research team manually reviewed the corpus. These expressions were semantically normalized by mapping them to standardized terms such as “认知训练” (cognitive training) or “娱乐” (entertainment) or removed if deemed ambiguous. This step aimed to reduce the noise in the biterm co-occurrence structure of the BTM model, thereby enhancing the consistency and stability of topic clustering. Finally, the Baidu stop word list was applied to remove common non-informative words, yielding the final processed text. By analyzing these online reviews, this study provides deeper insights into users’ real needs, preferences, and experiences with elderly-friendly cognitive products.

Based on Equations (1) and (2) and using the Gibbs sampling technique, iterative calculations were performed for topic numbers ranging from 1 to 10. For each topic number, the model was run 10 times to evaluate its performance under different settings. Corresponding perplexity and coherence scores were calculated for each topic number. To facilitate a performance comparison across topic counts, both metrics were normalized, and a standardized trend graph was generated, as shown in Figure 8. The results show that as the number of topics increased, perplexity exhibited an overall monotonic decreasing trend, from 1.00 at K = 1 to 0.04 at K = 10, indicating improved model fit with more topics. However, relying solely on perplexity to determine the optimal number of topics can lead to semantic redundancy and overfitting. Therefore, coherence was introduced as a supplementary evaluation metric. When the number of topics was below 3, the interpretability of the generated phrases was poor. When the number exceeded 6, perplexity continued to decrease linearly, suggesting potential overfitting. Thus, the optimal topic range was between 3 and 6. Within this range, when K = 5, perplexity showed a clear inflection point and coherence reached its peak value (1.00), indicating that the generated topics are most cohesive, semantically distinct, and well clustered. Considering both metrics, K = 5 is identified as the optimal number of topics in this study, achieving a balance between semantic clarity and model fit.

In this study, the top 12 words with the highest probability for each topic were extracted using the BTM model. As these topics consisted of automatically generated word combinations without clear definitions, they could not directly inform design practices. Therefore, a focus group interpreted the phrases by transforming them into user needs and conducted a comprehensive analysis of the characteristic words and associated textual content. This process ultimately led to the identification of five key dimensions of user needs: cognitive training methods and effects, product appearance, leisure and entertainment, simplicity and convenience, and comfort and safety, as shown in

Table 3. Topic feature words and clusters for elderly-friendly game products.

Topics	Feature Words	Topic Cluster Name
Topic 1	coordination, inclusive, practical, gradual, progressive, puzzle, frequent, space, effect, coherent, universal, graded	Cognitive training methods and effects (X1)
Topic 2	vivid, nostalgic, simple, modern, minimalist, bright, intuitive, contrast, aesthetic, feature, classic, traditional	Product appearance (X2)
Topic 3	attractive, fun, satisfying, engaged, novel, imaginative, happy, interactive, changing, intergenerational, combination, encouraging	Leisure and entertainment (X3)
Topic 4	simple, compact, smooth, relaxed, skilled, lightweight, eye-catching, real-time, standardized, sensitive, dynamic, detailed	Simplicity and convenience (X4)
Topic 5	health, safety, comfort, friendly, sturdy, durable, ergonomic, easy to clean, anti-aging, non-toxic, odorless, easy to assemble	Comfort and safety (X5)

.

Through focus group analysis of the above topic identification results and characteristic words combined with aging-friendly and game product design, detailed user demand mapping was performed, and 23 specific needs for elderly-friendly puzzle products were identified, as shown in

Table 4. Mapping and analysis of user needs.

Topics	Feature Word Classification	User Needs	Need Description
Cognitive training methods and effects (X1)	inclusive, universal, practical	universality (X11)	Applicability and practicality of the product among the elderly user group
	frequent	long-term and high frequency (X12)	Increasing the frequency of product usage
	puzzle, space, coordination	cognitive training (X13)	Stimulate thinking and problem-solving skills to maintain the interest and cognitive ability of elderly users
	gradual, graded, progressive	different levels of difficulty (X14)	Products should have varying levels of difficulty
	effect, coherent	trackable progress (X15)	Provides detailed progress logging and analysis to help users and caregivers track results
Product appearance (X2)	vivid, bright, contrast, aesthetic	color coordination (X21)	Color schemes that cater to the aesthetics of elderly users
	simple, intuitive, modern, feature, minimalist	simplified appearance (X22)	Focus on functionality in overall design
	nostalgic, classic, traditional	nostalgic style (X23)	Increase elderly users’ interest and enjoyment in the product
Leisure and entertainment (X3)	changing, combination	rich gameplay (X31)	Provide flexible and diverse gameplay and enhance game challenges to maintain long-term attraction
	engaged, imaginative	immersive simulation (X32)	Create realistic gaming environments for enhanced engagement and control
	novel, attractive, fun	fun factor (X33)	Emphasize the entertainment and relaxation aspect to spark elderly user interest
	happy, satisfying	emotional satisfaction (X34)	Meet emotional care and psychological comfort
	encouraging	reward mechanism (X35)	Encourage sustained participation and offer a sense of achievement
	interactive, intergenerational	social entertainment (X36)	Foster interaction and communication among family members
Simplicity and convenience (X4)	real-time, sensitive, dynamic	feedback (X41)	Provide immediate and clear feedback to help elderly users understand their performance
	lightweight, compact	portability (X42)	Save space and facilitate easy carrying
	eye-catching, detailed, standardized	standardized instructions (X43)	Provide clear operation and mastery guidance
	smooth, simple, relaxed, skilled	easy operation (X44)	Simplify game product rules and operations
Comfort and safety (X5)	non-toxic, odorless, safety	material safety (X51)	Safe and health-friendly materials
	easy to clean, easy to assemble, anti-aging	easy maintenance (X52)	Ensure products maintain optimal performance
	sturdy, durable	durability (X53)	Improve product reliability and longevity
	comfort, ergonomic, friendly, health	comfort (X54)	Prioritize elderly users’ physiological characteristics

.

3.2. Analysis of the Weight of User Demand for Elderly Game Products

Based on the extraction of user needs outlined in Table 3 and adhering to the principles of AHP model construction, a hierarchical analysis model was developed, as shown in Figure 9.

After establishing the evaluation matrix, a focus group was invited to complete the assessment to ensure the scoring accuracy. Each level of need was compared and rated on a scale of 1 to 9. To construct a valid judgment matrix, AHP questionnaire responses were manually reviewed, and no significant outliers deviating from the mean were found. Therefore, the average method was applied to aggregate the scores for each comparison item, and the mean scores were calculated to derive the matrices in Equations (17)–(22).

$$X=\left[\begin{array}{ccccc}1& 2& {\displaystyle \frac{1}{2}}& {\displaystyle \frac{1}{2}}& 2\\ {\displaystyle \frac{1}{2}}& 1& {\displaystyle \frac{1}{3}}& {\displaystyle \frac{1}{5}}& {\displaystyle \frac{1}{2}}\\ 2& 3& 1& {\displaystyle \frac{1}{3}}& 3\\ 2& 5& 3& 1& 3\\ {\displaystyle \frac{1}{2}}& 2& {\displaystyle \frac{1}{3}}& {\displaystyle \frac{1}{3}}& 1\end{array}\right]$$

(17)

$$X_1=\left[\begin{array}{ccccc}1& 5& 2& 3& 5\\ {\displaystyle \frac{1}{5}}& 1& {\displaystyle \frac{1}{3}}& {\displaystyle \frac{1}{2}}& 1\\ {\displaystyle \frac{1}{2}}& 3& 1& {\displaystyle \frac{1}{2}}& 5\\ {\displaystyle \frac{1}{3}}& 2& 2& 1& 5\\ {\displaystyle \frac{1}{5}}& 1& {\displaystyle \frac{1}{5}}& {\displaystyle \frac{1}{5}}& 1\end{array}\right]$$

(18)

$$X_2=\left[\begin{array}{ccc}1& {\displaystyle \frac{1}{2}}& 2\\ 2& 1& 3\\ {\displaystyle \frac{1}{2}}& {\displaystyle \frac{1}{3}}& 1\end{array}\right]$$

(19)

$$X_3=\left[\begin{array}{cccccc}1& {\displaystyle \frac{1}{3}}& {\displaystyle \frac{1}{5}}& {\displaystyle \frac{1}{2}}& 2& 3\\ 3& 1& {\displaystyle \frac{1}{3}}& 2& 3& 5\\ 5& 3& 1& 3& 5& 5\\ 2& {\displaystyle \frac{1}{2}}& {\displaystyle \frac{1}{3}}& 1& 2& 3\\ {\displaystyle \frac{1}{2}}& {\displaystyle \frac{1}{3}}& {\displaystyle \frac{1}{5}}& {\displaystyle \frac{1}{2}}& 1& 1\\ {\displaystyle \frac{1}{3}}& {\displaystyle \frac{1}{5}}& {\displaystyle \frac{1}{5}}& {\displaystyle \frac{1}{3}}& 1& 1\end{array}\right]$$

(20)

$$X_4=\left[\begin{array}{llll}1& 3& {\displaystyle \frac{1}{2}}& {\displaystyle \frac{1}{3}}\\ {\displaystyle \frac{1}{3}}& 1& {\displaystyle \frac{1}{2}}& {\displaystyle \frac{1}{3}}\\ 2& 2& 1& {\displaystyle \frac{1}{2}}\\ 3& 3& 2& 1\end{array}\right]$$

(21)

$$X_5=\left[\begin{array}{llll}1& 2& {\displaystyle \frac{1}{2}}& {\displaystyle \frac{1}{3}}\\ {\displaystyle \frac{1}{2}}& 1& {\displaystyle \frac{1}{2}}& {\displaystyle \frac{1}{2}}\\ 2& 2& 1& {\displaystyle \frac{1}{2}}\\ 3& 2& 2& 1\end{array}\right]$$

(22)

To validate the logical consistency of the judgment matrix and reliability of the results, weights were calculated using the geometric mean method. A consistency check was performed, and the results are listed in

Table 5. Consistency check results.

Items	X	X1	X2	X3	X4	X5
λmax	5.166	5.236	3.009	6.202	4.168	4.145
CI	0.041	0.059	0.005	0.040	0.056	0.048
RI	1.120	1.120	0.520	1.260	0.890	0.890
CR	0.037	0.053	0.009	0.032	0.063	0.054

. All the CR values were below 0.1, confirming the feasibility of the calculations.

To determine the comprehensive weight of each sub-criterion, this study applied a multiplicative calculation method that integrated the weights of each criterion layer with their corresponding sub-criterion weights. The resulting comprehensive weights are presented in

Table 6. Weight values of comprehensive judgment matrix for user needs.

Criterion Layer	Weight	Sub-Criterion Layer	Weight	Comprehensive Weight	Ranking
X1	0.168	X11	0.419	0.0704	5
		X12	0.078	0.0131	20
		X13	0.209	0.0351	11
		X14	0.233	0.0391	9
		X15	0.061	0.0102	22
X2	0.072	X21	0.297	0.0214	15
		X22	0.539	0.0388	10
		X23	0.164	0.0118	21
X3	0.243	X31	0.102	0.0248	14
		X32	0.225	0.0547	6
		X33	0.411	0.0999	3
		X34	0.143	0.0347	12
		X35	0.065	0.0158	17
		X36	0.054	0.0131	19
X4	0.407	X41	0.192	0.0781	4
		X42	0.111	0.0452	8
		X43	0.255	0.1038	2
		X44	0.422	0.1718	1
X5	0.110	X51	0.177	0.0195	16
		X52	0.140	0.0154	18
		X53	0.264	0.0290	13
		X54	0.419	0.0461	7

.

According to the AHP weight results, the primary themes identified in this study were simplicity and convenience (X₄, 0.407), leisure and entertainment (X₃, 0.243), and cognitive training methods and effects (X₁, 0.168), followed by comfort and safety (X₅, 0.110) and product appearance (X₂, 0.072). In the subsequent design phase, the two highest-ranking user needs for each theme were selected to guide product development. Specifically, these included easy operation (X₄₄, 0.422) and standardized instructions (X₄₃, 0.255); fun factor (X₃₃, 0.411) and immersive simulation (X₃₂, 0.225); universality (X₁₁, 0.419) and different levels of difficulty (X₁₄, 0.233); comfort (X₅₄, 0.419) and durability (X₅₃, 0.264); and simplified appearance (X₂₂, 0.539) and color coordination (X₂₁, 0.297). In addition to these, two critical requirements, feedback (X₄₁, 0.0781) and portability (X₄₂, 0.0452), were identified, which ranked among the top 10 user needs based on comprehensive weight. Therefore, beyond the top two needs of each theme, these two dimensions should also be considered as key priorities in the subsequent product design process.

3.3. Technical Parameter Mapping of Elderly Game Products

Based on the key user needs of ease of use, fun, and comfort mentioned above, the information independence principle was combined to effectively map elements across different fields. This process transformed the elderly users’ requirement parameters (CA) into functional parameters (FA) for elderly-friendly game products, as shown in

Table 7. Key user needs and functional requirements for elderly-friendly puzzle game products.

User Needs (CAS)	Functional Requirements (FRS)
Universality CA11	Easily understandable information FR11
Different difficulty levels CA12	Selectable game difficulty FR12
Color coordination CA21	Bright and harmonious color scheme FR21
Simplified appearance CA22	Simple geometric shapes FR22
Immersive simulation CA31	Physically interactive products FR31
Fun factor CA32	Recreational game elements FR32
Feedback CA41	Immediate feedback response FR41
Portability CA42	Compact and lightweight FR42
Standardized instructions CA43	Clear product operation guide FR43
Easy operation CA44	Simple game operation FR44
Durability CA51	Durable materials FR51
Comfort CA52	Ergonomic dimensions suitable for elderly FR52

. The resulting matrix of the functional requirements is given by Equation (23).

$$\left\{\begin{array}{l}F{R}_{11}\\ F{R}_{12}\\ F{R}_{21}\\ F{R}_{22}\\ F{R}_{31}\\ F{R}_{32}\\ F{R}_{41}\\ F{R}_{42}\\ F{R}_{43}\\ F{R}_{44}\\ F{R}_{51}\\ F{R}_{52}\end{array}\right\}=\left\{\begin{array}{c}\mathrm{Easily} \mathrm{understandable} \mathrm{information}\\ \mathrm{Selectable} \mathrm{game} \mathrm{difficulty}\\ \mathrm{B}\mathrm{r}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t} \mathrm{a}\mathrm{n}\mathrm{d} \mathrm{h}\mathrm{a}\mathrm{r}\mathrm{m}\mathrm{o}\mathrm{n}\mathrm{i}\mathrm{o}\mathrm{u}\mathrm{s} \mathrm{c}\mathrm{o}\mathrm{l}\mathrm{o}\mathrm{r} \mathrm{s}\mathrm{c}\mathrm{h}\mathrm{e}\mathrm{m}\mathrm{e}\\ \mathrm{Simple} \mathrm{geometric} \mathrm{shapes}\\ \mathrm{Physically} \mathrm{interactive} \mathrm{products}\\ \mathrm{Recreational} \mathrm{game} \mathrm{elements}\\ \mathrm{Immediate} \mathrm{feedback} \mathrm{response}\\ \mathrm{Compact} \mathrm{and} \mathrm{lightweight}\\ \mathrm{Clear} \mathrm{product} \mathrm{operation} \mathrm{guide}\\ \mathrm{Simple} \mathrm{game} \mathrm{operation}\\ \mathrm{Durable} \mathrm{materials}\\ \mathrm{Ergonomic} \mathrm{dimensions} \mathrm{suitable} \mathrm{for} \mathrm{elderly}\end{array}\right\}$$

(23)

In alignment with aging-friendly design principles, the functional requirements of aging-friendly game products were further mapped into specific design parameters suitable for production and processing. The results are listed in

Table 8. Functional requirements and design parameters for elderly users of puzzle game products.

Functional Requirements (FRS)	Design Parameters (DPs)
Easily understandable information FR11	Common universal icons and symbols to reduce language dependency DP11
Selectable game difficulty FR12	Personalized game difficulty options DP12
Bright and harmonious color scheme FR21	Bright low-saturation primary color, with local high-saturation color DP21
Simple geometric shapes FR22	Circular, square, and triangular-shaped outlines DP22
Physically interactive products FR31	Modular physical props DP31
Recreational game elements FR32	Timed challenge games DP32
Immediate feedback response FR41	“Ding” sound cue for feedback DP41
Compact and lightweight FR42	Compact design, small and easy to move DP42
Clear product operation guide FR43	Add lighting guidance to the product interface to guide users to complete the game DP43
Simple game operation FR44	Single-hand-operated props DP44
Durable materials FR51	ABS plastic material DP51
Ergonomic dimensions suitable for elderly FR52	Enlarged icons and buttons based on elderly users’ visual and hand size characteristics DP52

and in the design parameter set matrix (24).

$$\left\{\begin{array}{c}D{P}_{11}\\ D{P}_{12}\\ D{P}_{21}\\ D{P}_{22}\\ D{P}_{31}\\ D{P}_{32}\\ D{P}_{41}\\ D{P}_{42}\\ D{P}_{43}\\ D{P}_{44}\\ D{P}_{51}\\ D{P}_{52}\end{array}\right\}=\left\{\begin{array}{c}\mathrm{Common} \mathrm{universal} \mathrm{icons} \mathrm{and} \mathrm{symbols} \mathrm{to} \mathrm{reduce} \mathrm{language} \mathrm{dependency}\\ \mathrm{Personalized} \mathrm{game} \mathrm{difficulty} \mathrm{options}\\ \mathrm{B}\mathrm{r}\mathrm{i}\mathrm{g}\mathrm{h}\mathrm{t} \mathrm{l}\mathrm{o}\mathrm{w}-\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{p}\mathrm{r}\mathrm{i}\mathrm{m}\mathrm{a}\mathrm{r}\mathrm{y} \mathrm{c}\mathrm{o}\mathrm{l}\mathrm{o}\mathrm{r},\mathrm{w}\mathrm{i}\mathrm{t}\mathrm{h} \mathrm{l}\mathrm{o}\mathrm{c}\mathrm{a}\mathrm{l} \mathrm{h}\mathrm{i}\mathrm{g}\mathrm{h}-\mathrm{s}\mathrm{a}\mathrm{t}\mathrm{u}\mathrm{r}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n} \mathrm{c}\mathrm{o}\mathrm{l}\mathrm{o}\mathrm{r}\\ \mathrm{Circular}, \mathrm{square}, \mathrm{and} \mathrm{triangular}-\mathrm{shaped} \mathrm{outlines}\\ \mathrm{Modular} \mathrm{physical} \mathrm{props}\\ \mathrm{Timed} \mathrm{challenge} \mathrm{games}\\ ”\mathrm{Ding}” \mathrm{sound} \mathrm{cue} \mathrm{for} \mathrm{feedback}\\ \mathrm{Compact} \mathrm{design}, \mathrm{small} \mathrm{and} \mathrm{easy} \mathrm{to} \mathrm{move}\\ \mathrm{Add} \mathrm{lighting} \mathrm{guidance} \mathrm{to} \mathrm{the} \mathrm{product} \mathrm{interface} \mathrm{to} \mathrm{guide} \mathrm{users} \mathrm{to} \mathrm{complete} \mathrm{the} \mathrm{game} \\ \mathrm{Sin}\mathrm{gle}-\mathrm{hand}-\mathrm{operated} \mathrm{props}\\ \mathrm{ABS} \mathrm{plastic} \mathrm{material}\\ \mathrm{E}\mathrm{n}\mathrm{l}\mathrm{a}\mathrm{r}\mathrm{g}\mathrm{e}\mathrm{d} \mathrm{i}\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{s} \mathrm{a}\mathrm{n}\mathrm{d} \mathrm{b}\mathrm{u}\mathrm{t}\mathrm{t}\mathrm{o}\mathrm{n}\mathrm{s} \mathrm{b}\mathrm{a}\mathrm{s}\mathrm{e}\mathrm{d} \mathrm{o}\mathrm{n} \mathrm{e}\mathrm{l}\mathrm{d}\mathrm{e}\mathrm{r}\mathrm{l}\mathrm{y} \mathrm{u}\mathrm{s}\mathrm{e}\mathrm{r}{\mathrm{s}}^{’}\mathrm{v}\mathrm{i}\mathrm{s}\mathrm{u}\mathrm{a}\mathrm{l} \mathrm{a}\mathrm{n}\mathrm{d} \mathrm{h}\mathrm{a}\mathrm{n}\mathrm{d} \mathrm{s}\mathrm{i}\mathrm{z}\mathrm{e} \mathrm{c}\mathrm{h}\mathrm{a}\mathrm{r}\mathrm{a}\mathrm{c}\mathrm{t}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{s}\mathrm{t}\mathrm{i}\mathrm{c}\mathrm{s}\end{array}\right\}$$

(24)

Incorporating the functional requirements and design parameters of elderly-friendly game products into Equation (9) results in the first-level design matrix for FRn and DPn, as shown in Equation (25). Based on the design matrix type shown in Figure 6, this matrix is a non-coupled diagonal matrix, confirming the rationality of the design parameters.

$$\left[\begin{array}{l}F{R}_1\\ F{R}_2\\ F{R}_3\\ F{R}_4\\ F{R}_5\end{array}\right]=\left[\begin{array}{lllll}X& O& O& O& O\\ O& X& O& O& O\\ O& O& X& O& O\\ O& O& O& X& O\\ O& O& O& O& X\end{array}\right]\left[\begin{array}{l}D{P}_1\\ D{P}_2\\ D{P}_3\\ D{P}_4\\ D{P}_5\end{array}\right]$$

(25)

A second-level design matrix was established for FRnm and DPnm. The final design matrix for elderly-friendly puzzle game products, as shown in Equation (26), is a non-coupled diagonal matrix. This conforms to the independence axiom of the AD theory, ensuring that the design parameters can align with user needs without conflict. Thus, these parameters were deemed reasonable and appropriate for subsequent manufacturing and processing.

$$\left[\begin{array}{l}F{R}_{11}\\ F{R}_{12}\\ F{R}_{21}\\ F{R}_{22}\\ F{R}_{31}\\ F{R}_{32}\\ F{R}_{41}\\ F{R}_{42}\\ F{R}_{43}\\ F{R}_{44}\\ F{R}_{51}\\ F{R}_{52}\end{array}\right]=\left[\begin{array}{llllllllllll}X& O& O& O& O& O& O& O& O& O& O& O\\ O& X& O& O& O& O& O& O& O& O& O& O\\ O& O& X& O& O& O& O& O& O& O& O& O\\ O& O& O& X& O& O& O& O& O& O& O& O\\ O& O& O& O& X& O& O& O& O& O& O& O\\ O& O& O& O& O& X& O& O& O& O& O& O\\ O& O& O& O& O& O& X& O& O& O& O& O\\ O& O& O& O& O& O& O& X& O& O& O& O\\ O& O& O& O& O& O& O& O& X& O& O& O\\ O& O& O& O& O& O& O& O& O& X& O& O\\ O& O& O& O& O& O& O& O& O& O& X& O\\ O& O& O& O& O& O& O& O& O& O& O& X\end{array}\right]\left[\begin{array}{l}D{P}_{11}\\ D{P}_{12}\\ D{P}_{21}\\ D{P}_{22}\\ D{P}_{31}\\ D{P}_{32}\\ D{P}_{41}\\ D{P}_{42}\\ D{P}_{43}\\ D{P}_{44}\\ D{P}_{51}\\ D{P}_{52}\end{array}\right]$$

(26)

3.4. Design and Evaluation of Elderly Game Products

Building on key design needs such as ease of operation, standardized instructions, and entertainment, technical parameters were established, including single-hand operation, lighting guidance, and timed challenge game rules. These parameters formed the basis for developing three design solutions for elderly-friendly game products, as illustrated in Figure 10.

Solution 1 focused on enhancing players’ hands-on skills, spatial thinking, and logical reasoning through simple assembly and combination. The props included small balls and variously shaped pipe modules, allowing players to connect freely and create different structures guided by lighting cues. This setup produced a dynamic effect as the small balls moved through the structure, fostering creativity and imagination. Each pipe connection used a magnetic structure to ensure smooth assembly, enabling easy one-handed operation for elderly users. The toy was made from safe and eco-friendly ABS plastic, offering durability and long-term safety. Additionally, a “skip function” button was included, allowing players to bypass difficult levels. The button featured a clear icon to enhance operational intuitiveness.

Solution 2 featured a simple rectangular design with core function buttons, including power and operation prompt buttons placed on one side. These buttons were paired with universally recognized icons to streamline operations and reduce product size, ensuring ease of use and portability for the elderly. The product panel provided lighting prompts to guide players and offered a clear and intuitive operational guide. The gameplay focused on spatial flipping using rotatable cube props that players should position correctly within a set time limit. The cube props were ergonomically designed to fit the hand dimensions of the elderly users and ensure comfort during use. This solution effectively trained spatial awareness and reaction abilities while emphasizing the simplicity of the prop operation.

Solution 3 incorporated the spatial flipping mechanism and introduced a two-player competitive mode to enhance interactivity. This feature fostered a sense of rivalry, motivating players to challenge themselves and increasing the game’s appeal. A “difficulty adjustment” button was included, allowing players to customize the game difficulty based on their skill level and cognitive abilities. Visually, the product employed bright color combinations to align with elderly users’ aesthetic preferences, serving as a visual guide to help players intuitively understand and operate the game, thereby reducing cognitive load. The props were designed using triangular geometric shapes to add novelty and increase the complexity of spatial operations. This design encouraged players to develop strong spatial thinking and operational skills, making it particularly effective for training hand–brain coordination while providing an enjoyable gaming experience.

This study assessed the three proposed design solutions along with the sample solution. The evaluation adhered to the TOPSIS calculation procedure outlined in Section 2.3.4, using the five topics derived from the BTM model as positive evaluation indicators. To further verify the effectiveness of the proposed solution in improving product acceptance, the existing “Jike” brand electronic building block puzzle toy in the market was selected as a sample solution, as shown in Figure 11. A panel comprising 26 elderly-friendly product design experts, 23 game product designers, 24 scholars in related fields, and 29 elderly users scored the solutions on a 7-point Likert scale. After standardizing the results, the initial evaluation matrix F was compiled, as shown in Equations (27)–(30). The relative closeness of each solution was calculated based on the scores of the four user groups.

$$F_1=\left[\begin{array}{lllll}5.1& 4.4& 4.5& 5& 4.5\\ 5.2& 4.3& 4.8& 5.1& 4.7\\ 4.8& 4.2& 4.6& 4.8& 4.6\\ 4.5& 4.1& 4.2& 3.9& 4.3\end{array}\right]$$

(27)

$$F_2=\left[\begin{array}{ccccc}4.6& 4.8& 4.8& 5.2& 4.9\\ 4.7& 4.9& 5& 5.2& 4.9\\ 4.8& 4.9& 5.1& 4.9& 4.8\\ 4.3& 4.3& 4.6& 4.4& 4.9\end{array}\right]$$

(28)

$$F_3=\left[\begin{array}{ccccc}5.2& 4.7& 4.9& 4.8& 4.8\\ 5.2& 4.8& 4.9& 5.1& 5\\ 5.2& 4.9& 4.8& 4.8& 4.9\\ 4.7& 4.6& 4.5& 3.9& 4.7\end{array}\right]$$

(29)

$$F_4=\left[\begin{array}{ccccc}5.2& 4.7& 4.6& 4.9& 4.4\\ 5.4& 5& 4.6& 4.9& 4.8\\ 4.9& 4.9& 4.7& 4.7& 4.7\\ 4.3& 4.2& 3.8& 4.2& 4.5\end{array}\right]$$

(30)

The relative closeness of each solution was calculated based on the scores of the four user groups. The results are shown in Figure 12.

Solution 2 had the highest relative closeness value, confirming that it was the optimal choice. Conversely, the existing market design ranked the lowest, demonstrating the reliability of this study’s design pathway for elderly-friendly game products. To verify the stability of the rankings generated by the TOPSIS method, 15 product designers and older adults were invited to jointly evaluate alternatives using the GRA method. Centered on the five thematic topics extracted from the BTM model and the design schemes, the solutions were evaluated using a 5-point Likert scale. In GRA, a higher grey relational grade indicates a better alternative. As shown in

Table 9. Comparison of ranking results between the TOPSIS method and the Grey Relational Analysis method.

Indicator	Closeness Coefficient (c)	Grey Relational Degree (r)
Design Solution 1	0.776	0.755
Design Solution 2	0.940	0.938
Design Solution 3	0.649	0.734
Sample solution	0.000	0.523
Solution ranking	Design Solution 2 > Design Solution 1 > Design Solution 3 > Sample solution	Design Solution 2 > Design Solution 1 > Design Solution 3 > Sample solution

, a comparison between the two methods revealed that solution 2 consistently ranked as the top solution, confirming the stability and reliability of the TOPSIS results.

To prevent resource waste during production and development, the optimal solution was further refined, as shown in Figure 13. The gameplay mechanism in Solution 2 involved controlling the rotation of a block within a defined space, with players navigating the block along a light-guided trajectory. The objective was to position the orange face of the block at the designated endpoint within a specified time frame. This design trains the players in terms of spatial and logical thinking. Upon winning, the system played a congratulatory sound to boost emotional engagement and encourage continued play. Players can use a single finger to control the movement of a block in four directions (up, down, left, and right), ensuring ease of use for elderly users. In case of errors, the system provides immediate feedback with a “ding” sound to help players identify and correct their actions. A detailed gameplay flowchart is shown in Figure 14. To enhance the gaming experience, block props can be modularly combined, increasing interactivity and difficulty in boosting engagement and enjoyment. Players can adjust their difficulty by rotating a dial, offering personalized options to suit various skill levels. Solution 2 was 20 cm long, 15 cm wide, and 2 cm high, making it compact and lightweight. The block props were designed as 2 × 2 cm cubes, an ergonomic size for elderly users to grip and operate. To improve safety, the product should have a simple geometric shape with rounded chamfers.

Many older adults experience high-frequency hearing loss, particularly reduced sensitivity to sounds above 4000 Hz [30]. According to ISO 7029, individuals over the age of 65 exhibit a significant decline in high-frequency auditory perception [31]. To enhance audibility, the product’s audio frequency range was set between 500 and 3000 Hz, covering the core range of human speech and better suited to elderly users. Regarding volume, a level of approximately 75 dB ensures clear auditory feedback while aligning with WHO recommendations for hearing safety in older adults [32], thereby improving product usability and experience across varying levels of hearing ability. In terms of materials, the upper portion of the product was made from durable ABS hard plastic. This choice considers the financial constraints of many older adults who rely on limited pension incomes, making material costs a key factor in product accessibility and adoption. A comparative analysis was conducted between two commonly used materials: ABS plastic and silicone compounds. Although silicone offers superior tactile comfort and anti-slip properties, its unit cost is significantly high. In contrast, ABS provides adequate rigidity and durability at a lower price, making it more appropriate for cost-sensitive elderly users. According to 2024 market data, the typical wholesale price of ABS ranges from USD 1.2 to 2.4 per kilogram [33], whereas Room Temperature Vulcanizing (RTV) silicone ranges from USD 10 to 20 per kilogram [34]. Additionally, silicone requires longer processing times, further increasing manufacturing costs. Even without considering processing differences, silicone’s unit cost can be two to three times that of ABS. Therefore, selecting ABS as the primary material without compromising basic functionality enhances economic feasibility and improves market accessibility for the elderly population. To maintain anti-slip performance, the product’s lower section, which occupies a smaller volume, incorporated firmer silicone while keeping overall costs controlled. In summary, the design not only improved gameplay through intuitive operation but also emphasized safety, convenience, and economic affordability, making it well suited for use by older adults.

4. Discussion

This study focused on aging-friendly game products and established a systematic design framework based on online review data mining. It aimed to address key challenges in the current design of game products for elderly users, such as the superficial identification of user needs, lack of targeted design solutions, and high subjectivity of evaluation processes. The results demonstrated that BTM topic modeling based on online reviews effectively extracted the core needs of elderly users, which primarily fell into five categories: simplicity and convenience, leisure and entertainment, cognitive training methods and effects, comfort and safety, and product appearance. These needs are largely driven by age-related cognitive and physical changes as well as the desire to enhance quality of life and self-confidence [35]. Previous research has shown that Chinese elderly users generally prioritize ease of use and practicality in aging-friendly products, particularly regarding operational convenience and functional utility [36], which aligns with the findings of this study. In traditional demand mining methods for aging-friendly designs, sources of user demand are primarily drawn from behavioral observations, user interviews, and questionnaire surveys [37]. These approaches often rely on assumed needs or expert evaluations and lack feedback derived from users’ long-term, real-world product experiences [38,39]. Therefore, this study leveraged actual consumer behavior as a key source of user demand and performed BTM topic modeling based on online shopping review texts from e-commerce platforms [40]. This approach enhances the authenticity of the data and the representativeness of user concerns, thereby providing a more accurate foundation for product design. It effectively captures user feedback and potential expectations regarding the gaming product experience, thereby addressing the limitations and one-sidedness of previous studies.

Second, the identified user needs were prioritized using AHP. The results showed that “simplicity and convenience” received the highest weight, indicating that operational burden remains the primary factor affecting product acceptance. This aligns with existing research, which highlights operational difficulty as a major barrier to user engagement [41]. Research has shown that in product design and development, priority should be given to reducing process complexity, streamlining operational steps, and simplifying information presentation to fulfill users’ preferences for a clear and straightforward experience [42]. Strategies such as “minimizing cognitive load” have been proposed to enhance task efficiency and product intuitiveness [43]. These findings highlight that “simplicity and ease of use” not only play a crucial role in user evaluation but also serve as a clear foundation for design decisions regarding function selection, structural layout, and interaction methods. This approach is particularly effective in reducing cognitive burden and operational barriers in aging-related products. Therefore, the results of this study provide direct and meaningful practical value for promoting usability-oriented design and development of products for the aging population. The second-highest priority was “leisure and entertainment”, which ranked above “cognitive training methods and effects”. This suggests that elderly users value the emotional and life-enhancing roles of game products, reflecting their growing importance in promoting mental health and well-being. This finding is consistent with previous studies showing that older adults are increasingly willing to engage in gaming to stimulate positive emotions and improve their mental health [44]. To meet the need for “leisure and entertainment”, aging-related product design can enhance user engagement by incorporating relaxing elements and playful features. For example, integrating point-based reward systems and modular tasks into functional design can enrich a product’s emotional value, extend its usage duration, and increase user retention [45]. Thus, the findings of this study contribute to strengthening the emotional connection between users and products while promoting long-term usage intentions [46]. They also offer clear guidance for functional planning and content hierarchy development throughout the product design process. Although “cognitive training methods and effects” did not rank as the top user priority, they remain essential for sustaining older adults’ long-term engagement and active interaction and therefore hold significant design value. Serious games have been shown to support cognitive intervention by improving core abilities such as memory and attention. This study emphasizes the importance of combining accessibility with progressive difficulty to accommodate the diverse backgrounds of elderly users. Progressive difficulty helps maintain cognitive challenges and user motivation through phased task modules, adaptive difficulty mechanisms, and real-time feedback. During product development, these user needs can be translated into key design parameters including game mechanics, interaction logic, and long-term engagement strategies. Popular game formats should aim to provide cognitive training that is inclusive across cultural and gender differences. The findings of this study offer a practical pathway for creating products that combine cognitive benefits with broad market adaptability.

In the comprehensive evaluation phase, this study further ranked the top 10 key user needs to ensure that the design process aligns with core user expectations. Notably, the weights of feedback and portability increased in composite scoring. This finding aligns with the established view that aging is accompanied by a decline in cognitive and physical abilities, particularly in managing operational complexity. Accordingly, providing feedback can help elderly users verify the correctness of their actions, thereby enhancing their sense of control and operational security [47]. Traditional research has primarily focused on the safety and ease of use in product design for the elderly [48] while paying relatively little attention to portability. However, this study revealed that older adults place a high value on product portability, which may be related to their desire to enhance their independence in daily life and reduce their risk of falling when carrying items. This finding addresses a limitation in the existing literature and underscores the need to strengthen the emphasis on portability in the design of elderly-focused products to support independent living. After collecting a large number of user needs, they must be translated into specific design solutions. However, existing studies often lack a structured mapping mechanism for translating user demands into concrete design parameters, resulting in limited relevance between identified needs and actual product functions [49]. To address this limitation, this study adopted a three-layer mapping model, “user demand–functional requirement–design parameter”, based on AD theory, thereby enhancing the practical operability of aging-friendly product design. Specifically, the identified user needs were systematically mapped to the corresponding design parameters using independence and information axioms. For example, cognitive-related needs were addressed through puzzle-based training in spatial and logical reasoning as well as through personalized selection of game difficulty levels. This structured approach ensured both scientific validity and practical feasibility of the design framework, culminating in the development of three aging-friendly game design proposals.

Finally, during the evaluation stage of the design schemes, the TOPSIS method was employed to conduct a comprehensive assessment. The results indicated that solution 2 was the optimal design, as it was closest to the positive ideal solution and farthest from the negative ideal solution, outperforming existing game products. This finding validates the effectiveness of the proposed design framework in demand mining, scheme generation, and optimization and further confirms the practical value of the core user concerns identified in this study as a guide for aging-friendly product design.

5. Conclusions

In conclusion, this study proposed a clearly structured and methodologically rigorous design process for elderly-friendly game products driven by data mining and integrating BTM, AHP, AD, and TOPSIS. Theoretically, this study addressed existing gaps in the extraction of elderly user needs, structured functional mapping, and multi-criteria objective evaluation. Practically, it enhanced the alignment between design solutions and the actual needs of elderly users. The proposed method not only is applicable to game product design but also shows potential for broader application in other elderly-friendly domains, such as terminal devices and rehabilitation aids, offering a replicable methodological path and theoretical support for the systematic design of aging-friendly products.

However, certain limitations of this study remain. This study focused primarily on game products for the elderly, and the applicability of the proposed framework to other product categories has not yet been empirically validated. Future research should broaden the scope of application to evaluate the generalizability of the framework across various types of elderly-oriented products. Additionally, this study emphasized the value of online review texts for user need extraction and therefore did not incorporate direct engagement methods such as interviews or observations with elderly users. Although this approach demonstrates the feasibility of using online reviews as a data source, it limits the comprehensiveness of user perspectives and the depth of design insights. To address this limitation, future studies should integrate qualitative methods such as interviews and field observations to gain deeper insights into elderly users’ behavioral patterns and emotional needs, thereby enhancing the model’s explanatory power and applicability. Moreover, the robustness of research results under real-world manufacturing conditions remains an important area for future research. It is recommended that future work pursue active collaboration with industry partners to support the development and practical implementation of product prototypes based on the proposed design framework. Feedback from these design practices can further enrich and refine the theoretical model, assess its feasibility and effectiveness in real-world design contexts, and ultimately contribute to the optimization and industrialization of age-friendly products.