The development of precise vehicle emission models is crucial for estimating vehicular exhaust emissions. Though measuring emissions using an on-board emissions measurement system can be promising, it is essential to improve the precision of emission rates (ERs) prediction through effective statistical methods. A novel framework of simple linear regression (SLR), support vector regression (SVR), and piecewise linear regression (PLR) approaches was employed to develop a speed-based emission model. In total, 30 trips data from six professional drivers were collected to understand the variability of tailpipe emissions. The developed SLR, SVR, and PLR models demonstrated high accuracy, as indicated by mean absolute percentage error (), root-mean-square error (), and coefficient of determination () values. PLR outperformed SLR, and SVR in predicting CO, CO HC and NO ERs. These models can be useful tools for policymakers to understand emissions in heterogeneous traffic conditions and develop appropriate solutions to improve air quality.

中文翻译：

---

使用基于回归的方法对现实世界的柴油车尾气排放进行建模

精确的车辆排放模型的开发对于估算车辆尾气排放至关重要。尽管使用机载排放测量系统测量排放前景广阔，但通过有效的统计方法提高排放率 (ER) 预测的精度至关重要。采用简单线性回归 (SLR)、支持向量回归 (SVR) 和分段线性回归 (PLR) 方法的新颖框架来开发基于速度的排放模型。总共收集了 6 名专业驾驶员的 30 次出行数据，以了解尾气排放的变化情况。开发的 SLR、SVR 和 PLR 模型表现出较高的准确性，如平均绝对百分比误差 ()、均方根误差 () 和决定系数 () 值所示。PLR 在预测 CO、CO HC 和 NO ER 方面优于 SLR 和 SVR。这些模型可以成为政策制定者了解异构交通条件下的排放并制定适当的解决方案以改善空气质量的有用工具。