New Energy Vehicle Decision-Making for Consumers: An IBULIQOWA Operator-Based DM Approach Considering Information Quality

Abstract

New energy vehicles (NEVs) have gained increasing favor among NEV consumers due to their dual advantages of “low cost” and “environmental friendliness.” In recent years, the share of NEVs in the global automotive market has been steadily rising. For instance, in the Chinese market, the sales of new energy vehicles in 2024 increased by 35.5% year-on-year, accounting for 70.5% of global NEV sales. However, as the diversity of NEV brands and models expands, selecting the most suitable model from a vast amount of information has become the primary challenge for NEV consumers. Although online service platforms offer extensive user reviews and rating data, the uncertainty, inconsistent quality, and sheer volume of this information pose significant challenges to decision-making for NEV consumers. Against this backdrop, leveraging the strengths of the quasi OWA (QOWA) operator in information aggregation and interval basic uncertain linguistic information (IBULI) information aggregation and two-dimensional information representation of “information + quality”, this study proposes a large-scale group data aggregation method for decision support based on the IBULIQOWA operator. This approach aims to assist consumers of new energy vehicles in making informed decisions from the perspective of information quality. Firstly, the quasi ordered weighted averaging (QOWA) operator on the unit interval is extended to the closed interval $\left[0,\tau \right]$, and the extended basic uncertain information quasi ordered weighted averaging (EBUIQOWA) operator is defined. Secondly, in order to aggregate groups of IBULI, based on the EBUIQOWA operator, the basic uncertain linguistic information QOWA (BULIQOWA) operator and the IBULIQOWA operator are proposed, and the monotonicity and degeneracy of the proposed operators are discussed. Finally, for the problem of product decision making in online service platforms, considering the credibility of information, a product decision-making method based on the IBULIQOWA operator is proposed, and its effectiveness and applicability are verified through a case study of NEV product decision making in a car online service platform, providing a reference for decision support in product ranking of online service platforms.

1. Introduction

Today, as global climate change intensifies and environmental protection becomes increasingly urgent, promoting the transformation to a green economy has become a shared goal for countries worldwide [1,2,3]. New energy vehicles (NEVs), as a paradigm of green mobility, represent not only a critical direction for global industrial restructuring but also a vital driver for advancing the sustainable development of the world economy [4]. With the growing enhancement of global environmental awareness and the deepening integration of sustainable development principles, NEVs have gained increasing favor among eco-conscious consumers, leading to a steadily expanding consumer market. In 2024, global sales of NEVs reached 18.236 million units, with China’s share of global NEV sales rising from 64.8% in 2023 to 70.5%. However, given the rapid development of the new energy vehicle market, NEV consumers are often confronted with a wide array of automobile brands [5,6]. Regarding research on the purchasing behavior of NEV consumers, scholars have conducted theoretical studies from various perspectives, including NEV-related policies, consumer cognition, and the sensitivity of purchase intentions [7]. At the level of practical application and game theory, researchers have introduced and integrated game theory into the analysis of consumer behavior during the purchase decision-making process [8], thereby validating the significance of incorporating consumer behavior into NEV-related decision making. Building upon this foundation, this study aims to start from the “information collection–purchase decision” phase of the consumer behavior cycle. It seeks to extract high-quality, decision-relevant content from multi-dimensional and large-scale user reviews across cross-platform online automotive service platforms. Based on this extraction, a comprehensive decision-making framework for NEV consumers is constructed, with the goal of assisting them in achieving a more satisfactory decision-making process during the “information collection–purchase decision” stage.

Meanwhile, the rapid convergence of new-generation information technologies such as cloud computing, mobile internet, and big data has led to an exponential growth in decision-making data across various fields. Consequently, big data has progressively transformed traditional model-driven decision research into a more data-driven approach. Furthermore, multi-criteria decision making (MCDM) research is also transitioning from being primarily model-driven to being predominantly driven by big data [9]. MCDM [10] is a crucial approach in decision making, which encompasses the comprehensive consideration of multiple criteria for selection of an optimal solution. The advent of big data has provided a robust foundation for research in MCDM, offering substantial data support for product categorization and recommendation studies across various domains such as automotive, tourism, and catering through the extensive availability of online service platforms [11,12]. For example, existing decision-making models based on traditional methods such as VIKOR [13], TOPSIS [14], and AHP [15] are primarily limited to ranking alternatives based on multi-source data. However, these models do not consider that the data used in decision-making experiments, often sourced from big data, may have quality issues—issues that can significantly affect subsequent decision outcomes. Thus, the presence of data quality issues presents challenges to research on big data-driven multi-criteria decision making, as inaccurate data can undermine the reliability and value of decision outcomes [16]. Therefore, examining the issue of information quality in decision-making research within the context of data-driven decision making holds innovative significance for enabling NEV consumers to make scientifically sound and rational product evaluations when selecting NEVs.

The integration of decision information constitutes a pivotal element within the decision making (DM) framework, particularly in the context of amalgamating group information in large-scale data-driven decision research, wherein aggregation functions assume an indispensable role. The ordered weighted averaging (OWA) [17] operator, being a fundamental operator in the aggregation function system, has emerged as a widely adopted tool for information integration within the DM method system due to its inherent ability to incorporate both the intrinsic importance of information and the significance of its positional sorting. In recent years, scholars have progressively expanded the application of the ordered weighted averaging operator to various types of fuzzy information environments [18], proposing a range of aggregation operators such as the intuitionistic fuzzy OWA operator [19] and Pythagorean fuzzy OWA operator [20]. Furthermore, scholars have proposed an enhanced limited set of OWA operators [18], which incorporate the unique feature group information frequently encountered in decision-making problems within complex scenarios, thereby expanding the practical application domains of OWA operators. The quasi OWA operator [20], an extension of the OWA operator, encompasses a range of aggregation operators including the OWA operator [17] and the ordered weighted geometric averaging (OWGA) operator [17]. Due to its robust fusion characteristics [21], extensive research has been conducted by scholars in this area. For instance, Liu [22] investigated orness measurement and application value of composite quasi OWA operators on closed intervals $\left[a,b\right]$. Furthermore, Yang and Chen [23] extended the quasi OWA operator to fuzzy environments, proposing intuitionistic fuzzy quasi OWA operators and applying them in product decision-making fields to demonstrate their versatility. The aforementioned research has highlighted the advantages of the OWA operator and its derivative aggregation methods in handling information characterized by ambiguity and uncertainty, as well as their broad applicability across various domains. Building upon these strengths, this paper proposes an aggregation framework designed to synthesize a large volume of user reviews from online automotive service platforms.

Meanwhile, considering trust and information quality in decision making has also become an important area for scholarly exploration. Scholars have addressed the issue of trust in people’s decision-making processes, through approaches such as community networks [24,25] and consensus reached [26,27], enriched the theoretical foundation of decision making from a trust-based perspective. However, despite incorporating trust as a key factor, these studies did not take into account the quality of the experimental data and information used. Meanwhile, decision-making models based on community discovery and consensus inevitably encounter fundamental challenges related to data quality. Therefore, focusing exclusively on data quality considerations, this paper sets aside the consensus, coordination, and interaction processes typical of traditional group decision-making models. Instead, it proposes a large-scale independent data aggregation approach that fundamentally differs from conventional group decision-making methods, with the aim of enriching the internal structure of the decision-making framework. Currently, to address the inherent uncertainties in decision making, scholars have conducted relevant research from various perspectives [28,29,30], including user rationality, consumption sentiment, and product feature extraction. However, in decision-making studies based on user evaluation data, the quality of the information itself has not been sufficiently integrated into the processes of information aggregation and decision making. Incorporating information quality into the process of information aggregation poses a significant challenge. In recent years, scholars have proposed the concept of basic uncertain information (BUI) [31,32], which characterizes information $x$ and its credibility $c$ through a binary tuple $〈x,c〉$. Due to its provision of a framework for representing information quality, it has gradually gained widespread attention from scholars. Additionally, a series of multi-criteria decision-making methods have been proposed. In the research on aggregation operators for BUI, scholars have proposed the basic uncertain information ordered weighted averaging (BUIOWA) operator [33] and the basic uncertain information ordered weighted geometric averaging (BUIOWGA) operator [34] in a finite set environment.

In complex and uncertain decision-making environments, decision-makers tend to employ a set of qualitative semantic terms to characterize preferences, thereby affording themselves greater flexibility in expressing evaluations of specific objects. Therefore, researchers have expanded BUI into a two-tuple linguistic term set [35] environment and introduced basic uncertain linguistic information [36] (BULI) and interval basic uncertain linguistic information (IBULI). Based on BUI and its derivative theories, scholars have explored the advantages of information quality. From the perspective of information aggregation, researchers have proposed aggregation operators that integrate consensus and rough set theory (as presented in two research papers). In the field of recommendation systems, a credibility-based engineering material decision-making ranking system [37] and a cross-platform automotive product ranking system have been introduced [38,39]. Furthermore, to reduce the subjectivity in attribute weight assignment within decision-support systems, scholars have proposed a weight determination method that synthesizes public and expert opinions from the perspective of information quality [40]. However, existing studies indicate that, from the perspective of information quality, the development of IBULI and its aggregation operators—capable of integrating semantic expression variations and decision-makers’ credibility—remains at an early stage. Therefore, this paper aims to integrate the QOWA operator’s capability in handling semantic generality with the broad applicability of information quality-based approaches, proposing an aggregation support model for information on big data evaluation that targets NEV consumer decision making while incorporating information quality considerations.

In conclusion, the aggregation of large-scale data plays a crucial role in decision support within big data-driven scenarios. Online service platform data constitutes a primary source of internet-oriented data. However, user comment data within such platforms often suffers from quality issues. For instance, in real-world contexts, there exists the phenomenon of “review manipulation”—a sudden surge of overly positive reviews. Despite this, few existing studies consider information quality when aggregating decision-making information. Neglecting such quality issues may inevitably lead to biased decision making by distorting the perception of products, thereby causing deviations in the outcomes generated by decision-making methodologies. To address this issue, this paper integrates the advantages of the IBULI representation method—which incorporates the dual dimensions of “information + credibility”—with the QOWA operator, known for its ability to detect information deviations effectively. Based on this integration, the IBULIQOWA operator is proposed, along with a corresponding ranking method, aiming to provide foundational support for information aggregation in the decision-making processes of NEV consumers.

2. Description of Decision Making for NEV Consumer in NEV and Information Quality Analysis

2.1. Description of Decision Making for NEV Consumer

With the growing awakening of consumer awareness, the NEV sector is experiencing a robust development trend, characterized by a continuous emergence of brands and products. In this context, the vehicle recommendations offered by online automotive service platforms have progressively become a crucial information source for assisting NEV consumers in making rational purchasing decisions. Figure 1 and Figure 2 respectively provide a summary overview of the ratings and user evaluations for the “Model Y” NEV on platforms. Specifically, Figure 1 visually presents the overall rating of the “Model Y” NEV, encompassing multi-dimensional quantitative evaluation indicators such as overall performance, endurance capacity, intelligent configuration, and more. This provides consumers with a comprehensive and objective reference framework for assessing product quality. Furthermore, it offers a ranking recommendation for similar NEV products using both the “average operator” and “aggregation operator” to assist target consumers in their decision-making process. In contrast, Figure 2 focuses on real user feedback, aggregating personal experiences and feelings from numerous consumers during actual use. These include detailed comments on driving performance, satisfaction with interior craftsmanship, and after-sales service satisfaction. Such user evaluation data reflect the market performance of the “Model Y” NEV and reveal consumer awareness and acceptance at a micro level. For potential consumers, these evaluations serve as crucial references and decision-support tools, aiding them in selecting products that better align with their needs and expectations.

Based on the analysis of user ratings, evaluation data from existing online automotive service platforms, and the NEV product recommendation mechanism, these platforms—by leveraging their extensive user interaction networks—can provide NEV consumers with rich decision-making information and product selection suggestions. However, in the current NEV product ranking process, platforms often rely solely on the simple arithmetic average aggregation method to process large-scale user rating data, which fails to adequately account for variations in the quality of user-provided information. Given the lack of a reliability assessment for such large-scale network user feedback data, this approach may produce recommendation outcomes that do not accurately reflect consumers’ preferences, thereby undermining their ability to make optimal decisions.

In light of the aforementioned limitations in the product recommendation processes of current online automotive service platforms, this paper proposes a decision-making model based on information quality. By integrating a user credibility evaluation algorithm and constructing a multi-dimensional information quality weighting system, the model enhances the ranking mechanism of recommendation outcomes. The objective is to provide a more scientific and adaptive decision-support framework for NEV consumers, thereby improving supply–demand matching efficiency in the NEV market and promoting the development of green consumption.

2.2. Information Processing and Quality Analysis

The primary focus of this study is to address the NEV decision-making problem within online automotive service platforms. Due to the vast territory of China, there are significant regional disparities in the purchasing power of new energy vehicles. A general aggregation of individual user data is not only complex but also inappropriate, as it may obscure regional differences and lead to decision-making outcomes that fail to accurately reflect local market conditions. To address this issue, this paper integrates data characteristics with China’s geographical regional features by subdividing the dataset according to regions and assigning appropriate regional weights accordingly. For the sake of convenience, we present the relevant formulas for product decision-making problems in a uniform manner as follows:

Alternative solution set: The set $U=\left\{u_i|i=1,2,\dots ,n\right\}$ consists of $n$ distinct categories of alternative automotives.

Regional division set: The set $z=\left\{1,2,\dots ,s\right\}$ denotes the ensemble of regional partitions for comprehensive evaluation, while $\mathit{\omega }=\left({\omega }_1,{\omega }_2,\dots ,{\omega }_s\right)$ is a weight vector that satisfies conditions ${\omega }_k\in \left[0,1\right]$ and ${\displaystyle \sum _{k=1}^s{\omega }_k}=1$.

Rating matrices set: The matrix $R_i={\left(r_i^y\right)}_{y\times 1}\left(i=1,2,\dots ,n\right)$ represents the comprehensive rating matrix of scheme $u_i\in U$, where $r_i^y\in \left[1,5\right]$ denotes the overall evaluation provided by user $y$ for scheme $u_i\in U$.

Based on the predefined set, this article will subsequently perform a quality analysis of the large-scale evaluation data obtained from the online automotive service platform. The analysis will focus on aspects such as user certification, content, and user attention index. We provide an illustrative example of user information and credibility indicators based on an online automotive service platform, as shown in Figure 3.

The process of information quality analysis involves the following steps:

Step 1: Acquire the submatrix of ratings based on regional segmentation.

The rating matrix set $R_i={\left(r_i^y\right)}_{y\times 1}$ of product $u_i\in U$ is partitioned into submatrices $R_i^z$ based on geographical regions (South China $z=1$, Central China $z=2$, North China $z=3$, West China $z=4$, and East China $z=5$).

$$\left\{\begin{array}{l}{R}_i={\cup }_{z=1}^5{R}_i^z,i=1.\dots ,n;z=1,2,\dots ,5\\ R_i^z={\left(r_i^{y_z}\right)}_{y_z\times 5},z=1,\dots 5\end{array}\right.$$

(1)

where $R_i^z$ represents a submatrix about the region $z$, and $y_z$ represents the total number of users in that area.

Step 2: Transform the rating submatrix $R_i^z$ into an interval 2-tuple linguistic information matrix $GI{R}_i^z={\left(I{R}_i^z\right)}_{1\times 1}$.

$$\begin{array}{l}I{R}_i^z=\left[{\mathrm{\Delta }}_S\left({\psi }_i^{z-}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{z+}\right)\right]\\ \hspace{1em}=\left[\begin{array}{l}{\displaystyle \frac{1}{2}}\cdot \left({\min }_{y_z}\left\{r_i^{y_z}-1\right\}+{\displaystyle \frac{1}{y_z}}\cdot {\displaystyle \sum _{y=1}^{y_z}\left(r_i^y-1\right)}\right),\\ {\displaystyle \frac{1}{2}}\cdot \left({\max }_{y_z}\left\{r_i^{y_z}-1\right\}+{\displaystyle \frac{1}{y_z}}\cdot {\displaystyle \sum _{y=1}^{y_z}\left(r_i^y-1\right)}\right)\end{array}\right]\end{array}$$

(2)

where ${\psi }_i^{z-},{\psi }_i^{z+}\in \left[0,4\right]$.

Step 3: Acquire the credibility matrix $CGI{R}_i^z={\left(c_i^z\right)}_{1\times 1}$ for submatrix $R_i^z$.

Firstly, it is imperative to acquire the credibility matrix $C{R}_i={\left(c{r}_i^y\right)}_{y\times 1}$ for all users corresponding to the rating matrix $R_i={\left(r_i^y\right)}_{y\times 1}$. The calculation procedure is outlined as follows:

Step 3.1: Calculate the user’s certification index, denoted as $G_1\left(r_i^y\right)$, based on the platform’s authentication of the rating user.

$$G_1\left(r_i^y\right)=\left\{\begin{array}{l}1,Certified users\\ 0.6,Uncertified users\end{array}\right.$$

(3)

Step 3.2: Calculate the user’s attention index $G_2\left(r_i^y\right)$ by considering the number of secondary comments $TL\left(r_i^y\right)$ received from user reviews.

$$G_2\left(r_i^y\right)=\left\{\begin{array}{l}1.0,TL\left(r_i^y\right)>25\\ 0.6,25\ge TL\left(r_i^y\right)>10\\ 0.2,TL\left(r_i^y\right)<10\end{array}\right.$$

(4)

Step 3.3: Determine the content index $G_3\left(r_i^y\right)$ for each user based on their assessed proficiency level on the service platform, taking into account user reviews.

$$G_3\left(r_i^y\right)=\left\{\begin{array}{l}1.0,\mathrm{Full}-\mathrm{recommend} \\ 0.6,R\mathrm{e}\mathrm{c}\mathrm{o}\mathrm{m}\mathrm{m}\mathrm{e}\mathrm{n}\mathrm{d} \\ 0.2,No-\mathrm{recommend} \end{array}\right.$$

(5)

Step 3.4: Calculate the user’s group consistency index $G_4\left(r_i^y\right)$ by assessing the similarity between their rating and the average rating of the group.

$$G_4\left(r_i^y\right)=1-{\displaystyle \frac{1}{y_z}}{\displaystyle \sum _{y_0=1}^{y_z}\left|\left(r_i^{y_o}-1\right)-\frac{1}{y_z}{\displaystyle \sum _{y_0=1}^{y_z}\left(r_i^{y_0}-1\right)}\right|}$$

(6)

Step 3.5: Synthesize the aforementioned four types of indices to derive a credibility matrix $C{R}_i^z={\left(c{r}_i^{y_z}\right)}_{y_z\times 5}$.

$$c{r}_i^y={\displaystyle \frac{1}{4}}\left(G_1\left(r_i^y\right)+G_2\left(r_i^y\right)+G_3\left(r_i^y\right)+G_4\left(r_i^y\right)\right)$$

(7)

Step 3.6: Acquire the credibility submatrix $CGI{R}_i^z={\left(c_i^z\right)}_{1\times 1}$ from the regional submatrix $GI{R}_i^z={\left(I{R}_i^z\right)}_{1\times 1}$.

$$c_i^z={\displaystyle \frac{1}{y_z}}{\displaystyle {\sum }_{y=1}^{y_z}c{r}_i^y}$$

(8)

Step 4: Integrate the region’s information matrix $GI{R}_i^z={\left(I{R}_i^z\right)}_{1\times 1}$ and credibility matrix $CGI{R}_i^z={\left(c_i^z\right)}_{1\times 1}$ into an IBULI matrix $GIB{R}_i={\left(IB{R}_i^z\right)}_{1\times 5}$.

$$IB{R}_i^z=〈I{R}_i^z;c_i^z〉=〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{z-}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{z+}\right)\right];c_i^z〉$$

(9)

where $IB{R}_i^z$ refers to the comprehensive evaluation data of IBULI products $u_i\in U$ user interface for users in region $z=1,2,3,4,5$.

3. Consumer Rating Information Aggregation Operator

The current section primarily focuses on the IBULIQOWA operator. It presents the relevant fundamental definitions to provide essential operator support for the NEV consumer DM model proposed in this paper.

3.1. IBULI

In relation to the interval basic uncertain linguistic information (IBULI), the concept of an interval 2-tuple linguistic variable is introduced as the first step.

Definition 1 ([37]).

Let $S=\left\{S_0,S_1,\mathrm{\dots },S_{\tau }\right\}$ be an ordered linguistic term set (LTS), then

(1)

The 2-tuple pair $A_i=〈\left[\left(S_{i^{-}},{\alpha }_i^{-}\right),\left(S_{i^{+}},{\alpha }_i^{+}\right)\right];c_i〉$ represents interval basic uncertain linguistic information (IBULI), which can be expressed as follows: $A_i=〈\left[{\psi }_i^{-},{\psi }_i^{+}\right];c_i〉$, where $\left[{\psi }_i^{-},{\psi }_i^{+}\right]\in \left[0,\tau \right],{\psi }_i^{-}\le {\psi }_i^{+}$, $c_i\in \left[0,1\right]$.

(2)

When ${\psi }_i^{-}={\psi }_i^{+}={\psi }_i$, $A_i=〈\left[{\psi }_i^{-},{\psi }_i^{+}\right];c_i〉=〈{\psi }_i;c_i〉$ induces to BULI [36].

(3)

The set $IB$ represents all pairs of binary vectors in the IBULI framework, and $〈{\mathrm{\Delta }}_s\left(\tilde{\mathit{\psi }}\right);\mathit{c}〉:\left[n\right]\to IB$ is an n-dimensional mapping function for IBULI vectors that satisfies condition

$$〈{\mathrm{\Delta }}_S\left(\tilde{\mathit{\psi }}\right);\mathit{c}〉={\left(〈{\mathrm{\Delta }}_S\left({\tilde{\psi }}_i\right),c_i〉\right)}_{i=1}^n\in I{B}^n$$

(10)

where the interval 2-tuple linguistic vector is denoted by ${\mathrm{\Delta }}_S\left(\tilde{\mathit{\psi }}\right):\left[n\right]\to \left[0,\tau \right]$, while the credibility vector is represented by $\mathit{c}:\left[n\right]\to \left[0,1\right]$.

(4)

Regarding the two mapping functions of IBULI, $I{P}_{\psi }:IB\to \left[0,\tau \right]$ and $I{P}_c:IB\to \left[0,1\right]$, they satisfy

$$I{P}_{\psi }\left(〈{\mathrm{\Delta }}_S\left({\tilde{\psi }}_i\right),c_i〉\right)={\mathrm{\Delta }}_S\left({\tilde{\psi }}_i\right),I{P}_c\left(〈{\mathrm{\Delta }}_S\left({\tilde{\psi }}_i\right),c_i〉\right)=c_i$$

(11)

For detailed definitions and theorems related to IBULI, please refer to Appendix A.

Definition 2 ([37]).

The pair of IBULIs, denoted as $A_i=〈{\mathrm{\Delta }}_S\left({\tilde{\psi }}_i\right),c_i〉$, is characterized by its score function $S\left(A_i\right)$ and hesitation degree $H\left(A_i\right)$, which are defined as follows:

$$S\left(A_i\right)=\left[{\displaystyle \frac{{\psi }_i^{-}+{\psi }_i^{+}}{2\tau }}\right]\cdot c_i,H\left(A_i\right)=\left[{\displaystyle \frac{{\psi }_i^{+}-{\psi }_i^{-}}{\tau }}\right]\cdot c_i$$

(12)

Definition 3 ([37]).

The ranking method between two IBULI pairs, denoted as $A_i=〈{\mathrm{\Delta }}_S\left({\tilde{\psi }}_i\right),c_i〉$ and $A_j=〈{\mathrm{\Delta }}_S\left({\tilde{\psi }}_j\right),c_j〉$, is defined as follows:

(1)

If $S\left(A_i\right)>S\left(A_j\right)$, then $A_i\succ A_j$, $A_i$ is superior to $A_j$;

(2)

If $S\left(A_i\right)=S\left(A_j\right)$, then

(1)

If $H\left(A_i\right)<H\left(A_j\right)$, we have $A_i\succ A_j$;

(2)

If $H\left(A_i\right)=H\left(A_j\right)$, then

(1)

Given that $c_i>c_j$, then $A_i\succ A_j$;

(2)

Given that $c_i=c_j$, then $A_i~A_j$.

where $S\left(A\right)$ and $H\left(A\right)$ represent the score function and hesitation function of IBULI, respectively.

3.2. IBULIQOWA Operator

In this subsection, we introduce the IBULIQOWA operator, which is designed to aggregate large-scale group data and support decision-making processes. Let $E$ be a finite set, where ${[0,1]}^E$ represents the set of operations defined on all $x:E\to [0,1]$ (or all fuzzy sets over $E$). The evaluation of $x:\left[n\right]\to [0,1]$ in any nonempty subset $E$ of $\left[n\right]$ can be denoted as $x{|}_E:E\to [0,1]$, and for any $i\in E$, we obtain $x{|}_E\left(i\right)=x\left(i\right)$. A simple and often encountered situation is that some input values (input vectors) are obtained with full uncertainty, and the remainder of input values (in that input vector) are obtained with full certainty. That is, for an input function/vector $\mathit{x}:\left[n\right]\to \left[0,1\right]$ and a crisp nonempty subset $E\subseteq \left[n\right]$, the input values in ${\left\{x\left(i\right)\right\}}_{i\subseteq E}$ are all with full certainty, and the input values in ${\left\{x\left(i\right)\right\}}_{i\subseteq \left[n\right]\backslash E}$ are all with full uncertainty.

Definition 4.

Let $〈{\mathrm{\Delta }}_S\left(\tilde{\mathit{\psi }}\right);\mathit{c}〉$ be a set of IBULI pairs, $E\subseteq \left[n\right]$ be a finite set (its complement $\left[n\right]\backslash E$), and the IBULIQOWA operator be denoted as mapping $IBULIQOW{A}_{\mathit{\omega }}:IEB{L}^n\to IEBL$, satisfying condition

$$\begin{array}{l}IBULIQOW{A}_{\mathit{\omega }}\left({\mathrm{\Delta }}_s\left(\tilde{\mathit{\psi }}\right);\mathit{c}\right)=IBULIQOW{A}_{\mathit{\omega }}\left({\mathrm{\Delta }}_s\left(\left[{\mathit{\psi }}^{-},{\mathit{\psi }}^{+}\right]\right);\mathit{c}\right)\\ =〈{\mathrm{\Delta }}_s\left(\left[\begin{array}{l}I{P}_{\mathit{\psi }}\left(EBUIQOW{A}_{\mathit{\omega }}\left(〈{\mathrm{\Delta }}_s\left({\mathit{\psi }}^{-}\right);\mathit{c}〉\right)\right),\\ I{P}_{\mathit{\psi }}\left(EBUIQOW{A}_{\mathit{\omega }}\left(〈{\mathrm{\Delta }}_s\left({\mathit{\psi }}^{+}\right);\mathit{c}〉\right)\right)\end{array}\right]\right),P_c\left(EBUIQOW{A}_{\mathit{\omega }}\left(〈\mathit{\psi };\mathit{c}〉\right)\right)〉\end{array}$$

(13)

The operators $EBUIQOW{A}_{\mathit{\omega }}\left(〈{\mathrm{\Delta }}_s\left({\mathit{\psi }}^{-}\right);\mathit{c}〉\right)$ and $EBUIQOW{A}_{\mathit{\omega }}\left(〈{\mathrm{\Delta }}_s\left({\mathit{\psi }}^{+}\right);\mathit{c}〉\right)$ are the BULIQOWA operators corresponding to $〈{\mathrm{\Delta }}_S\left({\mathit{\psi }}^{-}\right);\mathit{c}〉$ and $〈{\mathrm{\Delta }}_S\left({\mathit{\psi }}^{+}\right);\mathit{c}〉$, respectively, where

$$\begin{array}{l}I{P}_{\mathit{\psi }}\left(EBUIQOW{A}_{\mathit{\omega }}\left(〈{\mathrm{\Delta }}_s\left(\mathit{\psi }\right);\mathit{c}〉\right)\right)\\ ={\displaystyle \underset{0}{\overset{1}{\int }}\frac{\left|E_t\right|}{n}\cdot M_{f,{\mathit{\omega }}^{\left(\left|E_t\right|\right)}}\left(\mathit{\psi }{|}_{E_t}\right)dt}+{\displaystyle \underset{0}{\overset{1}{\int }}\left(1-\frac{\left|E_t\right|}{n}\right)\cdot Mea{n}_{f,\left[n\right]\backslash E_t}\left(\mathit{\psi }{|}_{\left[n\right]\backslash E_t}\right)}\end{array}$$

(14)

$$E{P}_c\left(EBUIQOW{A}_{\mathit{\omega }}\left(〈\mathit{\psi };\mathit{c}〉\right)\right)={\displaystyle \underset{0}{\overset{1}{\int }}AC\left(E_t\right)dt={\displaystyle \underset{0}{\overset{1}{\int }}\left(\left|E_t\right|/n\right)dt=\left(1/n\right){\displaystyle \sum _{i=1}^n\mathit{c}\left(i\right)}}}$$

(15)

The theorem and its proof are shown in Appendix B.

4. An IBULIQOWA Operator-Based DM Approach

In this chapter, we will develop a methodology that involves converting ratings to IBULI and integrating it with the IBULIQOWA operator to effectively address the product DM issue on the online automotive service platform. Figure 4 illustrates the process description of this study, focusing on the purchase decision-making issue of new energy vehicles for NEV consumers.

4.1. A Decision-Making Model Based on the IBULIQOWA Operator

Step 1: Input the IBULI matrix $GIB{R}_i={\left(IB{R}_i^z\right)}_{1\times 5}$ acquired in Section 2.2.

$$IB{R}_i^z=〈I{R}_i^z;c_i^z〉=〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{z-}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{z+}\right)\right];c_i^z〉$$

(16)

Step 2: Aggregate the regional IBULI matrices into a comprehensive IBULI matrix $IB{R}_i={\left(IB{R}_i\right)}_{1\times 1}$.

Step 2.1: Specify the attitude parameter value $\delta ={\mathrm{\Omega }}_{f,\mathit{\omega }}$ and solve for regional aggregate weights using the subsequent model.

$$\begin{array}{l}\mathrm{Objective} \mathrm{function}:\min \left|{\mathrm{\Omega }}_{f,\mathit{\omega }}-\delta \right|\\ s.t\left\{\begin{array}{l}{\mathrm{\Omega }}_{f,\mathit{\omega }}={\displaystyle \frac{f^{-1}\left({\displaystyle {\sum }_{i=1}^{n}f\left(0+\frac{n-i}{n-1}\left(\tau -0\right)\right){\omega }_i}\right)-0}{\tau -0}}\\ {\displaystyle \sum _{i=1}^n{\omega }_i}=1,{\omega }_i\in \left[0,1\right],i=1,2,\dots ,n\end{array}\right.\end{array}$$

(17)

where ${\mathrm{\Omega }}_{f,\mathit{\omega }}$ is the measure of orness in the IBULIQOWA operator, $f\left(x\right)=x^3$ is the generating function of the quasi mean operator, and $f^{-1}\left(x\right)=\sqrt[3]{x}$ is the inverse function of $f\left(x\right)$.

Step 2.2: Employ the IBULIQOWA operator to derive the comprehensive IBULI matrix $IB{R}_i={\left(IB{R}_i\right)}_{1\times 1}$.

$$\begin{array}{l}IB{R}_i=IBULIQOW{A}_{\mathit{\omega }}\left(IB{R}_i^1,IB{R}_i^2,\dots ,IB{R}_i^5\right)\\ =IBULIQOW{A}_{\mathit{\omega }}\left(\left(I{R}_i^1;c_i^1\right),\left(I{R}_i^2;c_i^2\right),\dots ,\left(I{R}_i^5;c_i^5\right)\right)\\ =BULIQOW{A}_{\mathit{\omega }}\left(\left(〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{1-}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{1+}\right)\right];c_i^1〉\right),\dots ,\left(〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{5-}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{5+}\right)\right];c_i^5〉\right)\right)\\ =〈{\mathrm{\Delta }}_s\left(\left[\begin{array}{l}{P}_{{\mathrm{\Delta }}_s\left({\mathit{\psi }}^{-}\right)}\left(EBUIQOW{A}_{\mathit{\omega }}\left(\left(〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{1^{-}}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{1^{+}}\right)\right];c_i^1〉\right),\dots ,\left(〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{5^{-}}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{5^{+}}\right)\right];c_i^5〉\right)\right)\right),\\ P_{{\mathrm{\Delta }}_s\left({\mathit{\psi }}^{+}\right)}\left(EBUIQOW{A}_{\mathit{\omega }}\left(\left(〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{1^{-}}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{1^{+}}\right)\right];c_i^1〉\right),\dots ,\left(〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{5^{-}}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{5^{+}}\right)\right];c_i^5〉\right)\right)\right)\end{array}\right]\right)\\ P_c\left(EBUIQOW{A}_{\mathit{\omega }}\left(\left(\left(〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{1^{-}}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{1^{+}}\right)\right]〉,c_i^1\right)\right),\dots ,\left(〈\left[{\mathrm{\Delta }}_S\left({\psi }_i^{5^{-}}\right),{\mathrm{\Delta }}_S\left({\psi }_i^{5^{+}}\right)\right]〉,c_i^5\right)\right)\right)\rangle \end{array}$$

(18)

Step 3: Obtain the sorted outcome of product $u_i\in U$ based on the ranking method for Definition 3.

$$u_{\sigma \left(i\right)}\succ u_{\sigma \left(i+1\right)},i=1,2,\dots ,n-1$$

(19)

4.2. An IBULIQOWA Operator-Based DM Approach for NEVs

The present section presents a decision-making method for ranking NEV products on an online vehicle service platform, based on the decision model constructed using the IBULIQOWA operator. The following specific steps are followed:

Phase 1: Identify the pertinent sets in the decision-making problem, namely, the NEV set of alternative solution $U=\left\{u_i|i=1,2,\dots ,n\right\}$, the set of regional division $z=\left\{1,2,\dots ,s\right\}$, and the set of rating matrices $R_i={\left(r_i^y\right)}_{y\times 1}\left(i=1,2,\dots ,n\right)$.

Phase 2: Apply the scoring and IBULI conversion method proposed in Section 2.2. to convert the rating matrix set $R_i={\left(r_i^y\right)}_{1\times y}$ into the IBULI matrix $GIB{R}_i={\left(IB{R}_i^z\right)}_{1\times 5}$.

Phase 3: Implement the decision model based on the IBULIQOWA operator proposed in Section 4.1. to derive the comprehensive IBULI matrix $IB{R}_i={\left(IB{R}_i\right)}_{1\times 1}$. Subsequently, employing the IBULI ranking method, obtain the sorted result of schemes $u_{\sigma \left(i\right)}\succ u_{\sigma \left(i+1\right)},i=1,2,\dots ,n-1$. The flowchart illustrating this product decision-making method is depicted in Figure 5.

5. Case Study

To demonstrate the proposed methodology in this study and validate its efficacy, we employed reputation data from the online automotive service platform Autohome and Dcar as a catalyst. We implemented the product decision-making approach suggested in this paper to derive the decision making regarding eight new energy vehicles and conducted a comparative analysis. To better reflect the practical implementation scenarios of the algorithm proposed in this paper, Figure 6 illustrates the flowcharts depicting user input parameters, data processing procedures, and result output mechanisms. The primary target users of the method proposed in this paper are NEV (new energy vehicle) consumers. In this decision-making ranking method based on large-scale data aggregation, NEV consumers are only required to input attitude parameters that reflect their personal decision-making preferences (determined by individual consumption preferences and financial capabilities). Specifically, users with a parameter value greater than 0.5 are classified as optimistic decision-makers, those equal to 0.5 as neutral consumers, and those less than 0.5 as pessimistic decision-makers. Subsequently, the system identifies vehicle user reviews from the crawled data and converts them into IBULI, an information representation integrating the dual dimensions of “information content” and “credibility.” The aggregated weight attributes are then calculated based on the attitude parameters provided by the user. Using the IBULIQOWA aggregation method, the system computes the comprehensive IBULI value for each piece of information, which reflects the overall score of each vehicle by combining user attitude preferences with information quality. Finally, by ranking these comprehensive values, the system outputs the name of the optimal car, representing the best decision-making option for users under their specified preferences and considering information’s reliability. Part of the running file code can be found in Supplementary Materials.

Phase 1: Determine crucial information, such as the alternative solution, regional division, and rating matrix.

NEV alternative solution set:

$$\begin{array}{l}U=\left\{u_i|i=1,2,\dots ,8\right\}\\ =\left\{Model3,ModelY,Han,Jikes001,LingpaoC11,QinPlus,\mathrm{Tang},XiaoPengP5\right\}\end{array}$$

Regional division set: $z=\left\{1,2,3,4,5\right\}$;

Table 1. The number of users and geographical information for NEV options (Autohome and Dcar).

Users	Model 3	Model Y	Han	JiKe001	LingPao C11	QinPLUS	Tang	XiaoPeng P5
Sum	536	549	804	664	779	987	589	873
$z=1$	132	107	246	274	252	411	262	245
$z=2$	180	209	130	114	110	111	78	134
$z=3$	78	84	183	106	142	182	52	267
$z=4$	70	72	68	93	98	114	65	110
$z=5$	66	77	177	77	177	169	122	117

presents the data volume associated with each car user.

Rating matrix set: $R_i={\left(r_i^y\right)}_{1\times y}\left(i=1,2,\dots ,8\right)$; refraining from presenting the rating information individually, this study acknowledges its extensive volume.

Phase 2: Employ the method proposed in Section 2.2. to perform a transformation on the rating matrix set $R_i$, resulting in the generation of the IBULI matrix $GIB{R}_i={\left(IB{R}_i^z\right)}_{1\times 5}$.

Step 1: Acquire the regional division-based rating submatrix $R_i^z\left(i=1,\dots ,8; z=1,2,\dots ,5\right)$. Owing to the extensive volume of data, a detailed representation of the rating submatrix is omitted

Step 2: Obtain the matrix of interval 2-tuple linguistic information $GI{R}_i^z\left(i=1,\dots ,8; z=1,\dots ,5\right)$, as elaborated in Appendix C.

Step 3: By calculating the authentication index $G_1\left(r_i^y\right)$, secondary comment index $G_2\left(r_i^y\right)$, content index $G_3\left(r_i^y\right)$, and group consistency index of all users, $G_4\left(r_i^y\right)$, we can derive the credibility matrix $C{R}_i\left(i=1,\dots ,8\right)$ for each user. Subsequently, we can obtain the corresponding submatrix from this credibility matrix.

$$CGI{R}_i^z\left(i=1,\dots ,8;z=1,\dots ,5\right)$$

(20)

See Appendix D.

Step 4: Obtain the IBULI matrix $GIB{R}_i\left(i=1,\dots ,8\right)$ for each integrated alternative solution, as elaborated in Appendix E.

Step 5: Consolidate the regional IBULI matrices into a comprehensive IBULI matrix $IB{R}_i\left(i=1,\dots ,8\right)$.

Step 5.1: Determine the regional aggregation weights by solving for $\delta =0.3$.

$$\begin{array}{l}{\omega }_1=\left(1\right);{\omega }_2=\left(0.3,0.7\right);{\omega }_3=\left(0.1,0.4,0.5\right)\\ {\omega }_4=\left(0.1,0.1,0.3,0.5\right);{\omega }_5=\left(0.1,0.1,0.1,0.2,0.5\right)\end{array}$$

(21)

Step 5.2: Obtain the comprehensive IBULI matrix $IB{R}_i\left(i=1,\dots ,8\right)$.

$$\begin{array}{cc}IB{R}_1=〈\left[2.4147,3.3943\right],0.566〉& IB{R}_2=〈\left[2.5607,3.519\right],0.5627〉\\ IB{R}_3=〈\left[2.8869,3.842\right],0.4315〉& IB{R}_4=〈\left[2.4848,3.3759\right],0.5043〉\\ IB{R}_5=〈\left[3.0155,3.8137\right],0.4452〉& IB{R}_6=〈\left[2.8431,3.8406\right],0.5177〉\\ IB{R}_7=〈\left[2.3933,2.8911\right],0.5088〉& IB{R}_8=〈\left[3.0984,3.8456\right],0.4537〉\end{array}$$

(22)

Phase 3: Obtain the NEV product $u_i\left(i=1,2,3\dots ,8\right)$ decision-making results in accordance with Definition 3, as illustrated in

Table 2. Scoring function and NEV alternative ranking results.

NEVs	Model 3	Model Y	Han	JiKe001	LingPao C11	QinPLUS	Tang	XiaoPeng P5
$\mathrm{Scoring} H_i$	0.41	0.43	0.36	0.37	0.38	0.43	0.34	0.39
Ranking	3	2	7	6	5	1	8	4

.

Then, the optimal NEV is $u_6$ (QinPLUS). To further and more thoroughly interpret the above experimental results,

Table 3. Ranking results of NEVs in five regions and averaging operator.

Regional Name	Ranking Results
South China	$u_6>u_2>u_1>u_8>u_5>u_4>u_3>u_7$
Central China	$u_6>u_2>u_7>u_1>u_4>u_8>u_5>u_3$
North China	$u_6>u_2>u_1>u_3>u_5>u_8>u_7>u_4$
West China	$u_2>u_6>u_7>u_4>u_5>u_3>u_8>u_1$
East China	$u_6>u_2>u_4>u_1>u_8>u_5>u_3>u_7$

presents the NEV ranking results from the perspective of each regional division and the result of the averaging operator. It can be observed that in the four regions of South China, Central China, North China, and East China, $u_6$ consistently ranks first. It ranks second only in West China, yet still remains among the top performers. This indicates that the optimal ranking results obtained by the method proposed in this paper hold significant value for the majority of users across the five major regions of the country.

6. Comparative Analysis

6.1. Orness Sensitivity Analysis

In step 2.1, decision-makers can determine the comprehensive evaluation BULI results that align with their personalized attitude preferences, based on their individual characteristics. To explore the impact of these preferences, a sensitivity analysis is conducted by designing different configurations of the decision-makers’ attitude parameter weights. Table 3 presents five scenarios of weight allocation. Specifically, orness values of 0 and 1 represent pessimistic and optimistic decision-making attitudes, respectively; an orness value of 0.5 indicates a neutral decision-making stance, while orness values of 0.3 and 0.7 reflect decision-making attitudes between pessimistic and neutral and between neutral and optimistic, respectively. Based on these five personalized weight scenarios,

Table 4. The result of orness sensitivity analysis.

Orness	Decision-Maker’s Attitude	Weight Vector	Optimal NEV
Orness = 0	Optimistic decision making	$\begin{array}{l}{\omega }_1=\left(1\right);{\omega }_2=\left(1,0\right);{\omega }_3=\left(1,0,0\right)\\ {\omega }_4=\left(1,0,0,0\right);{\omega }_5=\left(1,0,0,0,0\right)\end{array}$	$u_3$
Orness = 0.3	Optimistic-leaning decision making	$\begin{array}{l}{\omega }_1=\left(1\right);{\omega }_2=\left(0.3,0.7\right);{\omega }_3=\left(0.1,0.4,0.5\right)\\ {\omega }_4=\left(0.1,0.1,0.3,0.5\right);{\omega }_5=\left(0.1,0.1,0.1,0.3,0.4\right)\end{array}$	$u_6$
Orness = 0.5	Neutral decision making	$\begin{array}{l}{\omega }_1=\left(1\right);{\omega }_2=\left(0,5,0.5\right);{\omega }_3=\left(0.1,0.3,0.6\right)\\ {\omega }_4=\left(0.1,0.2,0.3,0.4\right);{\omega }_5=\left(0.1,0.1,0.1,0.2,0.5\right)\end{array}$	$u_6$
Orness = 0.7	Pessimistic-leaning decision making	$\begin{array}{l}{\omega }_1=\left(1\right);{\omega }_2=\left(0.7,0.3\right);{\omega }_3=\left(0.4,0.4,0.1\right)\\ {\omega }_4=\left(0.5,0.3,0.1,0.1\right);{\omega }_5=\left(0.5,0.2,0.1,0.1,0.1\right)\end{array}$	$u_7$
Orness = 1	Pessimistic decision making	$\begin{array}{l}{\omega }_1=\left(1\right);{\omega }_2=\left(0,1\right);{\omega }_3=\left(0,0,1\right)\\ {\omega }_4=\left(0,0,0,1\right);{\omega }_5=\left(0,0,0,0,1\right)\end{array}$	$u_8$

also displays the comprehensive ranking results of each NEV.

The results demonstrate significant variations in the ranking outcomes under different decision-maker attitude parameters. This suggests that the decision-making model proposed in this paper is capable of capturing the optimal decision-making outcomes of diverse consumers across all span parameters, thereby exhibiting broad applicability and flexibility in response to varying consumer attitudes. When decision-makers are optimistic, $u_3$ represents their optimal decision. However, if decision-makers are neutral or moderately optimistic, $u_6$ becomes the optimal choice. When decision-makers exhibit moderate pessimism or full pessimism, $u_7$ and $u_8$, respectively, represent the optimal decision-making strategies.

6.2. Theoretical Comparative Analysis

In the field of MCDM, traditional decision-making methods such as TOPSIS, VIKOR, and AHP have been widely applied to the ranking of product plans. Specifically, TOPSIS ranks alternatives based on their proximity to an idealized reference point; VIKOR is a multi-attribute decision-making method that simultaneously considers the maximization of group utility and the minimization of individual regret, while also incorporating decision-makers’ subjective preferences; and AHP is a method that performs qualitative and quantitative analyses by decomposing decision-related elements into hierarchical structures such as goals, criteria, and alternatives.

However, when dealing with data from online automotive service platforms characterized by significant variations in data quality, these traditional methods exhibit notable limitations. User evaluation information on such platforms often involves uncertainty, inconsistency, and substantial differences in quality, which traditional approaches struggle to handle effectively. In contrast, the fuzzy MCDM method, by integrating fuzzy set theory, can partially address the shortcomings of conventional decision-making techniques. Nevertheless, this method still lacks a systematic mechanism for evaluating information quality, which limits its applicability in complex environments with heterogeneous information quality. Against this background, this study proposes a large-scale group data aggregation method based on the IBULI framework, which demonstrates distinct advantages in addressing such challenges.

From the perspective of information quality-driven analysis, traditional MCDM methods are solely dependent on evaluation data for analysis of decisions, often overlooking the reliability of the information itself. In contrast, the IBULI model employed in this study enables dynamic evaluation of the quality of multi-source data. It is capable of not only representing the information content but also quantifying its reliability. By integrating information quality into the decision-making process, this approach enhances the reliability and robustness of decisions, thereby facilitating more scientifically sound and rational decision outcomes.

Table 5. A comparative analysis of different MCDM methods.

Methods	Data Applicability	Uncertainty	Information Quality	Ranking Result
TOPSIS	Structural data			$u_5>u_1>u_6>u_2>u_8>u_4>u_3>u_7$
VIKOR	Structural data			$u_1>u_6>u_5>u_8>u_2>u_4>u_3>u_7$
AHP	Hierarchical structure data			$u_6>u_5>u_8>u_1>u_2>u_4>u_3>u_7$
Averaging operator	Structural data			$u_6>u_5>u_2>u_1>u_8>u_4>u_3>u_7$
Fuzzy MCDM	Fuzzy data	√		$u_8>u_1>u_6>u_5>u_2>u_4>u_3>u_7$
Proposed method	Multi-source heterogeneous data	√	√	$u_6>u_2>u_1>u_8>u_5>u_4>u_3>u_7$

√ indicates that the method incorporates these factors.

presents the comparative analysis results of the method proposed in this paper against other MCDM methods. Meanwhile, according to the ranking results presented in the table, the method proposed in this paper demonstrates stability in the optimal decision $u_6$ when compared with other MCDM methods. Furthermore, in contrast to traditional MCDM methods that neglect information quality, the proposed approach exhibits reduced sensitivity to variations in information quality, thereby providing a more comprehensive and reliable ranking of products from the perspective of information quality.

7. Conclusions and Limitations

7.1. Conclusions

NEVs, as a critical domain for sustainable economic development, exhibit market performances intrinsically linked to NEV consumers’ decisions. These performances significantly influence both the NEV consumption market and the overall advancement of the sustainable economy. Given the current landscape characterized by numerous NEV brands and inconsistent data quality, NEV consumers encounter complex decision-making challenges. Leveraging a research context shaped by big data, this paper proposes a novel decision-making approach integrating the IBULIQOWA operator with the IBULI information representation method. This integrated approach facilitates the simultaneous consideration of information content and its inherent quality. By utilizing comprehensive user ratings and evaluation data sourced from large-scale automotive service platforms, we comprehensively assessed eight popular NEV models currently available on the market. The findings of this study not only assist NEV consumers in making more informed purchase decisions and advancing the adoption of sustainable consumption principles but also hold significant implications for fostering the development of sustainable NEV consumption patterns and the broader progression of the sustainable economy. Specifically, the contributions of this research can be summarized in both the realms of sustainable development and theory as follows:

(1)

Sustainable development perspective: This paper proposes a data-driven decision-making model that incorporates information quality into the decision-making process of NEV consumers regarding new energy vehicles. In the experimental section, using new energy vehicles as a case study, the decision-making mechanism of the model is thoroughly explained, highlighting how it supports NEV consumers’ choices. The feasibility and advantages of the proposed approach are demonstrated through comparative analysis with existing models, thereby confirming its effectiveness in significantly reducing decision-making costs and facilitating satisfactory outcomes for NEV consumers. Furthermore, at the managerial level, the data collected and the aggregated results provide multi-dimensional insights into consumers’ green purchasing preferences. These insights enable enterprises to leverage data analytics tools to better understand market trends and develop targeted marketing strategies, ultimately contributing to long-term sustainable development.

(2)

Theoretical perspective: This paper aims to address the issue of unreliable user rating information on online service platforms. To achieve this, the IBULI model is employed as a language representation framework for transforming decision-making criteria into structured information, which incorporates two key dimensions: rating information and information credibility. Furthermore, based on the OWA aggregation operator, the IBULIQOWA aggregation operator is developed to effectively aggregate IBULI-based information. Building on this foundation, a product ranking method is proposed to assist NEV consumers in selecting new energy vehicle products, thereby providing users with expected, reliable, and rational ranking outcomes.

These findings also provide actionable insights for policymakers, platform operators, and industry stakeholders aiming to promote sustainable consumption and mobility.

(1)

For policymakers: Transparent, data-driven decision-support tools can effectively guide consumer adoption of NEVs. Governments should establish unified standards for review disclosure across online platforms, thereby reducing information asymmetry. Furthermore, integrating user-generated data analytics into policy evaluation can help identify key consumer concerns—such as charging infrastructure and battery reliability—and enable the optimization of incentive schemes. These measures would accelerate the diffusion of NEVs and support national carbon neutrality objectives.

(2)

For platform operators: Online platforms play a crucial role in shaping consumer perceptions. To enhance the review system, platform operators should implement mechanisms to detect duplicate content, strengthen authenticity verification, and offer personalized recommendation features that accommodate diverse consumer preferences—such as those oriented toward sustainability versus price sensitivity. These enhancements would build consumer trust and promote informed, sustainability-driven purchasing decisions.

(3)

For industry stakeholders: Manufacturers and service providers can utilize aggregated consumer feedback to enhance product design, optimize after-sales service, and advance sustainability initiatives, such as battery recycling and energy efficiency improvements. By integrating consumer sentiment into experimental development and marketing strategies, industry practices can be better aligned with sustainability objectives, thereby improving long-term competitiveness.

7.2. Limitations

Nevertheless, it is imperative to conduct further investigations in future research to address the existing limitations.

(1)

Domain specificity: The current method is specifically designed for the NEV market. The unique characteristics of NEVs—such as strong policy support and rapid technological innovation—may not be fully applicable to other consumer product categories. Therefore, the generalizability of the findings is somewhat limited. Future research should apply the framework to other industries (e.g., home appliances, renewable energy equipment) to assess its broader applicability and robustness.

(2)

Cross-platform data integration: This study utilizes multi-modal data from Autohome and Dcar but aggregates them in a relatively simplistic manner. This approach overlooks potential cross-platform interactions, such as duplicated or overlapping reviews, which may bias the ranking outcomes and compromise the stability of the results. Future research should therefore develop more advanced data fusion techniques that explicitly account for interactivity and redundancy across platforms, thereby enhancing the validity and reliability of the findings.

(3)

Decision-maker heterogeneity: Although the paper examines various attitude parameter settings, unobserved heterogeneity in consumer decision making—such as differences in risk aversion, environmental concerns, or cost sensitivity—may still influence the stability of the ranking results. To enhance the robustness and adaptability of these rankings, future research should consider more granular classifications of decision-maker attitudes.