Objective The selection of automotive seat materials involves multiple heterogeneous indicators, including engineering performance, user perception, and environmental attributes, which are difficult to evaluate comprehensively using conventional methods. In addition, traditional weighting approaches rely heavily on expert judgment, leading to strong subjectivity and instability in the evaluation results. To address these challenges, this study aims to establish a data-driven comprehensive evaluation system for automotive seat materials by integrating large-scale user semantic information with objective engineering performance indicators.
Methods First, large-scale online user reviews related to automotive seat materials were collected from mainstream automotive discussion platforms. After data preprocessing procedures, including noise removal, word segmentation, and stop-word filtering, a high-quality text corpus was constructed. The Latent Dirichlet Allocation (LDA) topic model was then employed to automatically extract users’ latent concern themes from the textual data. To ensure both statistical robustness and semantic interpretability, the optimal number of topics was determined through a combined strategy involving perplexity analysis, LDAvis interactive visualization, and manual semantic validation. Based on the extracted user concern themes, a multi-level evaluation index system was established, covering three primary dimensions: functional performance, product attributes, and user experience. The construction of the index system was further refined by referencing relevant national standards, industry specifications, and expert knowledge, thereby enhancing its scientific rigor and practical relevance. Subsequently, the Fuzzy Analytic Hierarchy Process (FAHP) was applied to construct fuzzy complementary judgment matrices to capture the inherent uncertainty and ambiguity in expert evaluations. To overcome the limitations of traditional manual consistency adjustment in FAHP, Particle Swarm Optimization (PSO) was introduced to dynamically optimize the consistency and weight distribution of the fuzzy judgment matrices. Through iterative optimization, PSO effectively reduced subjective bias, improved matrix consistency, and enhanced the stability and reliability of the derived weight coefficients. Finally, three representative automotive seat materials—polyvinyl chloride (PVC) artificial leather, polyurethane (PU) artificial leather, and PU microfiber leather—were selected as case samples to validate the effectiveness and applicability of the proposed LDA−FAHP−PSO evaluation framework.
Results The topic modeling results indicate that users’ concerns regarding automotive seat materials can be mainly categorized into three core dimensions: durability and structural support, safety and environmental stability, and comfort with sensory experience. Comprehensive evaluation results demonstrate that PU microfiber leather achieves the highest overall score among the three materials. This superiority is primarily attributed to its outstanding performance in high-weight indicators, including wear resistance, structural support, aging resistance, safety, and environmental friendliness. PU artificial leather exhibits relatively balanced performance across multiple dimensions, particularly in terms of comfort, tactile sensation, breathability, and environmental attributes, indicating its strong competitiveness in the mid-range automotive market. In contrast, although PVC artificial leather possesses a clear cost advantage, its deficiencies in durability, aging resistance, and environmental performance significantly limit its competitiveness in high-quality and long-life applications. Notably, the evaluation outcomes show a high degree of consistency with the current market positioning and application scenarios of automotive seat materials, which further confirms the validity and rationality of the proposed evaluation model.
Conclusion This study proposed a comprehensive LDA−FAHP−PSO evaluation framework that could effectively integrate user semantic information with engineering performance indicators for automotive seat material selection. The proposed system successfully mitigated the subjectivity and instability inherent in traditional weighting methods and provided a reliable and quantitative decision-support tool for the scientific selection and optimization of automotive seat leather and alternative materials. Furthermore, the proposed framework demonstrated strong universality and scalability, and it could be readily extended to the intelligent evaluation of other automotive interior materials and complex multi-attribute composite material systems.
DownLoad: