Scholarlink

<p>As consumer preferences increasingly focus on emotional experiences, product design must address both functional and emotional needs. Traditional methods for collecting consumer feedback, such as surveys and focus groups, have limitations in terms of time and representativeness. Online reviews, however, offer valuable and real-time insights into consumer satisfaction and product perceptions. This study enhances product design by integrating machine learning with Kansei engineering, linking emotional responses directly to product features. To improve both the accuracy and interpretability of the model, we propose a multi-strategy improved Teaching–Learning-Based Optimization (TLBO) algorithm, optimized for the MITLBO-LightGBM model. In addition, the SHAP method is used to provide clear explanations of the model’s predictions, helping designers understand how specific design features impact consumer evaluations. By analyzing online review data, we identify the key product attributes that influence consumer satisfaction, offering a practical framework for data-driven product design improvements. </p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A more efficient and transparent product evaluation system: based on an improved LightGBM algorithm and interpretable machine learning methods

  • 期刊论文

  • 发表日期：

    2025年9月15日

查看全文
  • Binhao Huang,
  • Jian LV,
  • Ligang Qiang

摘要

As consumer preferences increasingly focus on emotional experiences, product design must address both functional and emotional needs. Traditional methods for collecting consumer feedback, such as surveys and focus groups, have limitations in terms of time and representativeness. Online reviews, however, offer valuable and real-time insights into consumer satisfaction and product perceptions. This study enhances product design by integrating machine learning with Kansei engineering, linking emotional responses directly to product features. To improve both the accuracy and interpretability of the model, we propose a multi-strategy improved Teaching–Learning-Based Optimization (TLBO) algorithm, optimized for the MITLBO-LightGBM model. In addition, the SHAP method is used to provide clear explanations of the model’s predictions, helping designers understand how specific design features impact consumer evaluations. By analyzing online review data, we identify the key product attributes that influence consumer satisfaction, offering a practical framework for data-driven product design improvements.