Scholarlink

<p>This study evaluates and compares the performance of traditional multiple linear regression (MLR) with advanced machine learning (ML) approaches such as support vector machine (SVM) and artificial neural network (ANN) for predicting household trip generation in Raipur, Chhattisgarh, India. Data was collected through household surveys comprising 1000 samples, incorporating socio-demographic characteristics, travel behaviour, and trip information. The analysis employed 16 independent variables including household characteristics and age categories, with daily trips as the dependent variable. After correlation analysis and variable selection (<i>p</i> &lt; 0.05), significant predictors were identified and used across all three models. The ANN model, with an 11-40-1 architecture, demonstrated superior performance (R<sup>2</sup> = 0.77, MSE = 1.048, MAE = 0.92), followed by MLR (R<sup>2</sup> = 0.686, MSE = 2.228, MAE = 1.18) and SVM (R<sup>2</sup> = 0.69, MSE = 2.21, MAE = 1.16). Total trip length and number of teenagers emerged as the most influential factors. While ANN showed the best prediction accuracy, MLR provided better interpretability, and SVM offered a balance between both. The results aid in understanding of trip generation trends in Indian cities and indicate that, although sophisticated machine learning methods can increase prediction accuracy, the modelling strategy selected should be in line with planning needs and the computational resources at hand..</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 Comparative Framework for Trip Generation Prediction: A Case Study of Raipur City, India

  • 期刊论文

  • 发表日期：

    2025年9月6日

查看全文
  • Saumya Anand,
  • Pritikana Das,
  • Bivina Geetha Rajendran

摘要

This study evaluates and compares the performance of traditional multiple linear regression (MLR) with advanced machine learning (ML) approaches such as support vector machine (SVM) and artificial neural network (ANN) for predicting household trip generation in Raipur, Chhattisgarh, India. Data was collected through household surveys comprising 1000 samples, incorporating socio-demographic characteristics, travel behaviour, and trip information. The analysis employed 16 independent variables including household characteristics and age categories, with daily trips as the dependent variable. After correlation analysis and variable selection (p < 0.05), significant predictors were identified and used across all three models. The ANN model, with an 11-40-1 architecture, demonstrated superior performance (R2 = 0.77, MSE = 1.048, MAE = 0.92), followed by MLR (R2 = 0.686, MSE = 2.228, MAE = 1.18) and SVM (R2 = 0.69, MSE = 2.21, MAE = 1.16). Total trip length and number of teenagers emerged as the most influential factors. While ANN showed the best prediction accuracy, MLR provided better interpretability, and SVM offered a balance between both. The results aid in understanding of trip generation trends in Indian cities and indicate that, although sophisticated machine learning methods can increase prediction accuracy, the modelling strategy selected should be in line with planning needs and the computational resources at hand..