Mathematics

Research

31 pages, 888 KB

Open AccessArticle

When Does Human–AI Collaboration Create Value in Live-Streaming Commerce?

by Yeyang Han and Ke Yan

Mathematics 2026, 14(15), 2817; https://doi.org/10.3390/math14152817 - 5 Aug 2026

Viewed by 222

Firms are increasingly introducing AI assistants into human-led live-streaming rooms, yet it remains unclear how such assistance should be configured and when it creates economic value. We develop an analytical model that compares human-only live-streaming selling with human–AI collaborative live-streaming selling. The firm [...] Read more.

Firms are increasingly introducing AI assistants into human-led live-streaming rooms, yet it remains unclear how such assistance should be configured and when it creates economic value. We develop an analytical model that compares human-only live-streaming selling with human–AI collaborative live-streaming selling. The firm sets the selling price in both modes and, under collaboration, jointly chooses the AI capability level. The model captures two channels through which AI may create value: enhancing the effectiveness of the human host and generating demand spillover beyond the room’s baseline conversion. We derive the equilibrium price, AI capability, demand, and profit, and we identify the conditions under which collaboration outperforms human-only live-streaming selling. The results show that AI capability is more valuable when paired with a stronger host, whereas the effect of product quality on AI investment is not necessarily positive. Lower AI cost may support a higher selling price by enabling a more capable selling process. Moreover, demand improvement and profit improvement need not occur simultaneously. Extensions examine AI-only live-streaming selling and imperfect AI assistance. Full article

(This article belongs to the Special Issue Advances in Artificial Intelligence, Machine Learning and Optimization, 2nd Edition)

► Show Figures

Figure 1

23 pages, 587 KB

Open AccessArticle

Artificial Intelligence Exposure, Perceived Job Replaceability, and Perceived Income Change: The Moderating Role of Task Codifiability—Evidence from the China General Social Survey

by Rong Nie, Xiaomei Bai and Jiangmin Ding

Mathematics 2026, 14(15), 2776; https://doi.org/10.3390/math14152776 - 3 Aug 2026

Viewed by 220

Abstract

This study examines how artificial intelligence (AI) exposure is associated with Chinese residents’ perceived income change and how this association varies with Perceived Job Replaceability. Using a nationally representative sample of 6247 employed workers from the 2021 China General Social Survey matched with industry-level [...] Read more.

This study examines how artificial intelligence (AI) exposure is associated with Chinese residents’ perceived income change and how this association varies with Perceived Job Replaceability. Using a nationally representative sample of 6247 employed workers from the 2021 China General Social Survey matched with industry-level robot penetration data, we estimate ordered probit models and two-stage residual-inclusion control-function checks within an ordered-response framework. Three findings emerge. First, objective AI exposure is associated with a 4.2-percentage-point lower probability of reporting higher household income than last year for a one-standard-deviation increase in robot density, based on predicted-probability contrasts rather than raw ordered-probit coefficients. Second, workers reporting low Perceived Job Replaceability are 15.7 percentage points more likely to report higher household income than last year on the same probability scale. Third, task codifiability moderates this relationship: marginal-effect contrasts show that low Perceived Job Replaceability is associated with a 22.9 percentage-point increase in the probability of reporting higher household income than last year in low-codifiability occupations, compared with only 4.7 percentage points in high-codifiability occupations, indicating a polarization pattern in perceived income change outcomes. Heterogeneity analyses further show that medium-skill routine workers face the strongest negative exposure associations, whereas high-skill workers in low-codifiability occupations show the strongest positive low-replaceability contrast. Overall, the findings clarify how AI exposure, worker perceptions, and task structure are jointly associated with perceived income change in China and provide evidence relevant to more inclusive technological adjustment policies. Full article

(This article belongs to the Special Issue Advances in Artificial Intelligence, Machine Learning and Optimization, 2nd Edition)

► Show Figures

Figure 1

29 pages, 5612 KB

Open AccessArticle

Rolling Bearing Fault Feature Extraction Based on Adaptive Hybrid Black-Winged Kite Optimized VME and SMHD

by Guanghe Zhu, Jiaqi Wang and Haijun Zhang

Mathematics 2026, 14(15), 2717; https://doi.org/10.3390/math14152717 - 31 Jul 2026

Viewed by 271

Abstract

Rolling bearing fault features are often weak and easily affected by noise and interference. To improve fault feature extraction performance, this paper proposes an AHBKA-VME-SMHD method. First, the black-winged kite algorithm is improved by opposition-based learning, a Gompertz-based adaptive step size strategy, and [...] Read more.

Rolling bearing fault features are often weak and easily affected by noise and interference. To improve fault feature extraction performance, this paper proposes an AHBKA-VME-SMHD method. First, the black-winged kite algorithm is improved by opposition-based learning, a Gompertz-based adaptive step size strategy, and an NGO-inspired random displacement strategy. Then, the improved algorithm is used to optimize the penalty factor and desired mode center frequency of VME, guided by a composite fitness function combining Higuchi fractal dimension and energy concentration index. Finally, SMHD is applied to enhance periodic impulsive components, and envelope spectrum analysis is used to identify fault characteristic frequencies. The proposed method is validated using simulated signals and two real-world bearing datasets, namely the CWRU and XJTU-SY datasets. The results show that the proposed method extracts clearer fault-related harmonics than the comparison methods. In addition, it obtains higher kurtosis and Gini index values and lower envelope spectrum entropy values, demonstrating its effectiveness for rolling bearing fault feature extraction. Full article

(This article belongs to the Special Issue Advances in Artificial Intelligence, Machine Learning and Optimization, 2nd Edition)

► Show Figures

Figure 1

20 pages, 1600 KB

Open AccessArticle

DiT1dLnet: A Fast and Accurate Diffusion Model Structure Based on Robot Behavior Imitation

by Jiaxin Liao, Weiyuan He, Qing Yu and Fei Chen

Mathematics 2026, 14(11), 1785; https://doi.org/10.3390/math14111785 - 22 May 2026

Viewed by 461

Abstract

A novel robot behavior generation method combining imitation learning with diffusion models elegantly addresses multi-modal action distributions, adapts to high-dimensional action spaces, and demonstrates impressive training stability. It significantly improves success rates across nine diverse tasks on three different robot simulation benchmarks, but [...] Read more.

A novel robot behavior generation method combining imitation learning with diffusion models elegantly addresses multi-modal action distributions, adapts to high-dimensional action spaces, and demonstrates impressive training stability. It significantly improves success rates across nine diverse tasks on three different robot simulation benchmarks, but comes with longer training times and slower inference speed. This paper proposes a novel architecture, DiT1dLnet, applied to DDPM for training and inference. DiT1dLnet improves accuracy across various robotic simulation tasks while accelerating training and inference speed by 50–100%. We benchmarked its performance on nine different tasks using three distinct robots. Full article

(This article belongs to the Special Issue Advances in Artificial Intelligence, Machine Learning and Optimization, 2nd Edition)

► Show Figures

Figure 1

21 pages, 1353 KB

Open AccessArticle

Causal-Patched Attention Network: Mitigating Contextual Bias and False Associations in Multi-Label Image Classification

by Baiqing Liu, Weiyuan He, Yingchang Jiang, Qing Yu and Fei Chen

Mathematics 2026, 14(9), 1521; https://doi.org/10.3390/math14091521 - 30 Apr 2026

Viewed by 474

Abstract

Multi-label image classification (MLIC) is vulnerable to contextual bias, where models may exploit spurious label–context associations rather than object evidence, leading to degraded generalization under distribution shifts. To address this issue, we propose CPAN, a causal-inspired framework that integrates label-specific feature decoupling, prototype-based [...] Read more.

Multi-label image classification (MLIC) is vulnerable to contextual bias, where models may exploit spurious label–context associations rather than object evidence, leading to degraded generalization under distribution shifts. To address this issue, we propose CPAN, a causal-inspired framework that integrates label-specific feature decoupling, prototype-based mediator modeling, patch-level evidence aggregation, and adaptive fusion. Specifically, CPAN uses a Transformer decoder to extract label-specific representations from the whole image and local patches. We introduce a prototype dictionary as a surrogate mediator space to encourage the model to rely on object-relevant intermediate patterns rather than context-sensitive shortcuts. We further aggregate patch-level predictions to enhance direct object evidence and fuse them with whole-image predictions through a learnable gate. Experiments on two benchmark datasets show that CPAN consistently improves both recognition accuracy and robustness. On MS-COCO, CPAN achieves 85.26 mAP, 80.67 CF1, and 82.52 OF1; on NUS-WIDE, it reaches 66.11 mAP, 64.42 CF1, and 75.95 OF1. Under context-shifted evaluation on MS-COCO, CPAN further obtains 80.93 mAP, 75.84 CF1, and 77.87 OF1, indicating stronger robustness to contextual bias. These results show that CPAN learns more object-centered representations and reduces reliance on spurious contextual correlations. Full article

(This article belongs to the Special Issue Advances in Artificial Intelligence, Machine Learning and Optimization, 2nd Edition)

► Show Figures

Figure 1

22 pages, 2200 KB

Open AccessArticle

A Novel K-Means with SHAP Feature Selection and ROA-Optimized SVM for Sleep Monitoring from Ballistocardiogram Signals

by Xu Wang, Fan-Yang Li, Yan Wang, Liang-Hung Wang, Wei-Yin Wu, Zne-Jung Lee, Wen Kang and Chien-Yu Lin

Mathematics 2026, 14(8), 1262; https://doi.org/10.3390/math14081262 - 10 Apr 2026

Viewed by 713

Abstract

Sleep quality is closely associated with cardiovascular, metabolic, and mental health outcomes, yet the clinical gold standard, polysomnography (PSG), is costly and intrusive for long-term home monitoring. Ballistocardiography (BCG) enables unobtrusive in-bed sensing and is therefore attractive for low-burden sleep assessment in natural [...] Read more.

Sleep quality is closely associated with cardiovascular, metabolic, and mental health outcomes, yet the clinical gold standard, polysomnography (PSG), is costly and intrusive for long-term home monitoring. Ballistocardiography (BCG) enables unobtrusive in-bed sensing and is therefore attractive for low-burden sleep assessment in natural environments. However, most existing BCG studies are PSG-referenced and mainly focus on sleep staging, while movement and out-of-bed episodes are often treated as artifacts rather than modeled jointly. In this study, we propose an interpretable unsupervised proxy-state modeling framework for three-state in-bed monitoring from BCG signals under an unlabeled setting. BCG recordings were segmented into 30 s windows with 50% overlap, and multi-domain features were extracted from waveform morphology, spectral power, heart rate-related dynamics, and wavelet energy distribution. K-means clustering (K = 3) was used to construct cluster-derived proxy labels, TreeSHAP-based feature ranking together with inner-CV-guided Top-N subset selection was used for training-only feature screening, and multiple classifiers were compared under a strict leave-one-subject-out protocol, with an ROA-optimized RBF-SVM achieving the best overall performance. Using data from 32 volunteers, the framework achieved an accuracy of 0.9932 ± 0.0047 (mean ± SD), together with consistently strong Macro-F1 and MCC scores. Overall, it outperformed the alternative methods compared in this study. Full article

(This article belongs to the Special Issue Advances in Artificial Intelligence, Machine Learning and Optimization, 2nd Edition)

► Show Figures

Figure 1

21 pages, 5426 KB

Open AccessArticle

Deep Learning-Based Recognition and Classification of Jin Cang Embroidery Stitches

by Ke-Ke Sun, Lu-Fei Yang, Zi-Ning Lan and Lu Gao

Mathematics 2026, 14(8), 1259; https://doi.org/10.3390/math14081259 - 10 Apr 2026

Viewed by 674

Abstract

Jin Cang embroidery, characterized by elaborate metallic threadwork and intricate textural patterns, is an important form of intangible cultural heritage. The digital preservation of Jin Cang embroidery is hindered by the scarcity of specialized datasets and the lack of object detection models that [...] Read more.

Jin Cang embroidery, characterized by elaborate metallic threadwork and intricate textural patterns, is an important form of intangible cultural heritage. The digital preservation of Jin Cang embroidery is hindered by the scarcity of specialized datasets and the lack of object detection models that balance high performance with computational efficiency for edge deployment. To address these challenges, a dedicated dataset comprising 3050 images across eight core stitch categories is introduced as the first dataset of its kind for Jin Cang embroidery. Building upon this foundation, Lite-YOLOv11s, a domain-specific lightweight detection framework, is proposed with MobileNetV4 as its backbone to improve the extraction of high-frequency texture cues associated with metallic threadwork. Experimental results show that Lite-YOLOv11s achieves an mAP@0.5 of 0.951, outperforming the YOLOv11s baseline (0.927) while reducing model parameters by 40% and FLOPs by 46%. EigenCAM visualizations further show that the model can localize discriminative stitch-level features even under complex backgrounds. This work provides an efficient and deployable solution for intelligent embroidery recognition and offers a useful reference for the digital preservation of other fine-grained cultural heritage crafts. Full article

(This article belongs to the Special Issue Advances in Artificial Intelligence, Machine Learning and Optimization, 2nd Edition)

► Show Figures

Figure 1

33 pages, 1613 KB

Open AccessFeature PaperArticle

Forecasting Risk Matrices with Economic Policy Uncertainty and Financial Stress: A Machine Learning Approach

by Jinda Du, Wenyi Cao and Ziyou Wang

Mathematics 2026, 14(6), 938; https://doi.org/10.3390/math14060938 - 10 Mar 2026

Cited by 2 | Viewed by 1199

Abstract

Accurately forecasting the risk matrix and constructing a well-controlled portfolio based on these forecasts is the core objective of effective asset allocation. This paper takes the Chinese stock market as the research object, employing multiple machine learning algorithms to systematically compare the predictive [...] Read more.

Accurately forecasting the risk matrix and constructing a well-controlled portfolio based on these forecasts is the core objective of effective asset allocation. This paper takes the Chinese stock market as the research object, employing multiple machine learning algorithms to systematically compare the predictive performance of the Financial Stress (FS) indicator and the Economic Policy Uncertainty (EPU) index in sectoral risk management. The forecast results are subsequently applied to portfolio construction and optimization. The findings indicate that, in terms of predictive dimensions, EPU demonstrates strong performance in short-term forecasts, but its explanatory power decays rapidly as the forecasting horizon extends. In contrast, the FS factor achieves forecasting accuracy that is significantly superior to both the EPU factor and traditional price series across all time horizons, exhibiting robust long-memory characteristics and cross-period stability. At the portfolio application level, the minimum variance strategy constructed based on FS forecasts effectively reduces out-of-sample portfolio variance, achieving superior risk control performance compared to strategies based on EPU factor forecasts. This result reveals the differentiated mechanisms of the two factor types: EPU acts as a driving force for short-term risk structure reshaping, while financial stress serves as the core variable driving the evolution of long-term risk structures. Machine learning methods provide an effective technical pathway for capturing these complex nonlinear relationships. The research conclusions offer new empirical evidence for investors to optimize asset allocation decisions and for regulatory authorities to improve risk monitoring systems. Full article

(This article belongs to the Special Issue Advances in Artificial Intelligence, Machine Learning and Optimization, 2nd Edition)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Advances in Artificial Intelligence, Machine Learning and Optimization, 2nd Edition

Share This Special Issue

Editors

Special Issue Information

Keywords

Benefits of Publishing in a Special Issue

Published Papers (8 papers)

Research

Further Information

Guidelines

MDPI Initiatives

Follow MDPI