For almost a century, several models have been developed to calibrate the pairwise relationship between traffic flow variables, that is, speed, density, and flow. Multi-regime models are well known for being superior over single-regime models in fitting the speed–density relationship. However, in modeling multi-regime models, breakpoints that separate the regimes are visually established based on the subjective judgment of data characteristics. Thus, this study proposes a data-driven approach to estimate the breakpoints of multi-regime models. It applies the Bayesian model for calibrating multi-regime models (two and three-regime models) for fitting traffic flow fundamental diagram. Furthermore, the analysis presented accounts for the random characteristics associated with the flow. To demonstrate the application of the proposed algorithm, traffic flow data from Interstate 10 (I-10) freeway in Jacksonville, Florida, were used in the analysis. The results demonstrate the potential benefit of using the proposed model in calibrating the fundamental diagram. The proposed approach can also quantify uncertainty and encode prior knowledge about the breakpoints in the model if the model developer wishes.

近一个世纪以来，已经开发了几种模型来校准交通流量变量之间的成对关系，即速度，密度和流量。众所周知，多区域模型在拟合速度-密度关系方面优于单区域模型。但是，在对多区域模型进行建模时，会根据数据特征的主观判断在视觉上建立分隔各个区域的断点。因此，本研究提出了一种数据驱动的方法来估计多区域模型的断点。它使用贝叶斯模型来校准多区域模型（两个和三个区域模型），以拟合交通流基本图。此外，提出的分析说明了与流量相关的随机特征。为了演示该算法的应用，分析中使用了佛罗里达州杰克逊维尔的10号州际公路（I-10）高速公路的交通流量数据。结果表明，使用所提出的模型校准基础图具有潜在的好处。如果模型开发人员愿意，建议的方法还可以量化不确定性并编码有关模型中断点的先验知识。