Scholarlink

<p>There are daily struggles for Bangladeshi citizens with an inefficient transportation system, but the experiences of university teachers and students are less researched. This study anticipates and investigates the major influencing factors on their commuting experience through novel machine learning-based feature selection and classification techniques. A systematic survey of 1,102 students of eight universities generated a tailored dataset, which was then analyzed with 16 feature selection methods (e.g., Fisher’s Score, Chi-square, L1/L2 regularization, genetic algorithm, ant colony optimization) and 15 classifiers (e.g., SVM, Random Forest, XGBoost, CatBoost, LGBM). Results indicate that 55.17% of the participants expressed dissatisfaction, and the most powerful drivers of dissatisfaction were found to be vehicle safety and social security. Non-linear SVM’s Fisher’s Score worked the best, and L2 Regularization happened the most frequently. Feature selection at higher levels greatly improved classification outcomes, highlighting its role in transport-oriented ML studies. Dataset and code are publicly accessible at <a href="https://kaggle.com/datasets/07e3786de7705329abd4ea9f2180271c278344ef759f26b188f48a1a951e501f">https://kaggle.com/datasets/07e3786de7705329abd4ea9f2180271c278344ef759f26b188f48a1a951e501f</a> and <a href="https://github.com/NahulRahman/Transport_Feature_Selection">https://github.com/NahulRahman/Transport_Feature_Selection</a>.</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

Discovering Key Influencing Factors of Transportation Mode Choice in Bangladeshi Universities Using Feature Selection Techniques

  • 期刊论文

  • 发表日期：

    2025年11月17日

查看全文
  • Md. Nahul Rahman,
  • Maisha Nanjeeba,
  • Nusrat Sharmin,
  • Sazia Tabasum Mim

摘要

There are daily struggles for Bangladeshi citizens with an inefficient transportation system, but the experiences of university teachers and students are less researched. This study anticipates and investigates the major influencing factors on their commuting experience through novel machine learning-based feature selection and classification techniques. A systematic survey of 1,102 students of eight universities generated a tailored dataset, which was then analyzed with 16 feature selection methods (e.g., Fisher’s Score, Chi-square, L1/L2 regularization, genetic algorithm, ant colony optimization) and 15 classifiers (e.g., SVM, Random Forest, XGBoost, CatBoost, LGBM). Results indicate that 55.17% of the participants expressed dissatisfaction, and the most powerful drivers of dissatisfaction were found to be vehicle safety and social security. Non-linear SVM’s Fisher’s Score worked the best, and L2 Regularization happened the most frequently. Feature selection at higher levels greatly improved classification outcomes, highlighting its role in transport-oriented ML studies. Dataset and code are publicly accessible at https://kaggle.com/datasets/07e3786de7705329abd4ea9f2180271c278344ef759f26b188f48a1a951e501f and https://github.com/NahulRahman/Transport_Feature_Selection.