Scholarlink

<p>Link performance function is a fundamental key input for the traffic assignment process. It depends on several parameters that can be calibrated to accommodate local conditions. However, the commonly used model is the Bureau of Public Roads (BPR) formula. This research attempts to calibrate this model parameters and develop new models with exponential regression analysis and a Multivariate Adaptive Regression Spline (MARS). Sufficient travel time data with the concurrent traffic flow were collected at different volume-to-capacity ratios, from urban arterials in Baghdad city. The resulting data points were used to calibrate the (BPR) model and develop new models. The ɑ and β parameters are calibrated as 0.8 and 4.7, respectively. The correlation coefficients of the calibrated models, the developed exponent model, and the developed MARS model are 0.83, 0.87, and 0.94, respectively, with a variance accounted of 0.78, 0.77, and 0.56, respectively. The MARS model, a kind of advanced data mining and machine learning approach, outperforms other models by saving approximately 35.6 and 27% of variance compared with the calibrated BPR model and exponent model, respectively. Selecting the most appropriate one depends on the method of analyzing the highway assignment and the available software. Since highway assignment is highly sensitive to link performance with varied traffic volume. </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

Data mining approach for calibrating and modeling the link performance function via MARS

  • 期刊论文

  • 发表日期：

    2025年7月14日

查看全文
  • Israa Saeed Al-Haydari,
  • Noor Moutaz Asmael

摘要

Link performance function is a fundamental key input for the traffic assignment process. It depends on several parameters that can be calibrated to accommodate local conditions. However, the commonly used model is the Bureau of Public Roads (BPR) formula. This research attempts to calibrate this model parameters and develop new models with exponential regression analysis and a Multivariate Adaptive Regression Spline (MARS). Sufficient travel time data with the concurrent traffic flow were collected at different volume-to-capacity ratios, from urban arterials in Baghdad city. The resulting data points were used to calibrate the (BPR) model and develop new models. The ɑ and β parameters are calibrated as 0.8 and 4.7, respectively. The correlation coefficients of the calibrated models, the developed exponent model, and the developed MARS model are 0.83, 0.87, and 0.94, respectively, with a variance accounted of 0.78, 0.77, and 0.56, respectively. The MARS model, a kind of advanced data mining and machine learning approach, outperforms other models by saving approximately 35.6 and 27% of variance compared with the calibrated BPR model and exponent model, respectively. Selecting the most appropriate one depends on the method of analyzing the highway assignment and the available software. Since highway assignment is highly sensitive to link performance with varied traffic volume.