Operational Analysis of Signalized Street Segments Using Multi-gene Genetic Programming and Functional Network Techniques

This article presents the operational analysis of urban signalized segments, the most fundamental entity of road networks, operating under heterogeneous traffic conditions. Geometric, traffic, and built-environmental data were collected from mid-block sections and downstream intersections of 45 well-diversified segments. Besides, the socio-demographic details, travel characteristics, and perceived satisfaction scores (varying from 1 $$=$$ excellent to 6 $$=$$ worst) were collected from 9000 on-street automobile drivers. Subsequently, the variables having significant ( $$p < 0.001$$ ) influences on the perceived satisfaction scores were identified by Spearman’s correlation analysis. As observed, the array of significant variables included six quantitative road attributes and the age group of motorists. By incorporating these variables, highly reliable but less complex automobile level of service (ALOS) models were developed with the help of two novel artificial intelligence techniques namely, multi-gene genetic programming (MGGP) and functional network (FN). Both models exhibited excellent prediction efficiencies in the present context and produced high coefficient of determination ( $$R^{2}$$ ) values of above 0.86 under the prevailing site conditions. The model comparison showed that the MGGP model is more reliable and easier for field implementations as compared to the FN model. The sensitivity analyses of modeled attributes revealed that traffic volume, travel speed, and automobile stop rate have by far the most significant influences on the ALOS of urban streets. The crucial outcomes of this study would largely help the transportation planners and engineers in quantifying the operational efficiencies of urban roadways and in taking efficient decisions for the better management of automobile traffic.