Development of a Modal Split Model Using Fuzzy Inference System

Abstract

Travel demand forecasting is four-stages procedure that includes trip generation, trip distribution, mode choice (Modal Split), and traffic assignment, it is an essential tool for predicting future demand. Among these, mode choice analysis is critical issue because it deals with forecasting the mode of transport utilized by travelers to reach their destination. The limitation of the traditional models used in mode choice analysis is that they do not take into account the subjectivity, imprecision, and ambiguity involved in human decisions. In this direction, fuzzy logic is found to be the most appropriate technique because it takes into account linguistic variables and expressions thus it attempts to harness human knowledge. This research was intended to develop modal split models by using fuzzy inference system and multiple linear regression. Ramadi city in Iraq was selected as a case study, this city is divided into 22 internal zones and 6 external zones, and a total of 28 traffic zones. The data related to this study was collected by home interview surveys, traffic survey for the entrances of Ramadi city, and from official departments. The socioeconomic characteristics (age, gender, income, car ownership, family size, trip cost, waiting time, and time spent inside the mode,) are selected as input variables. The results showed the fuzzy logic have a high prediction accuracy compared to the results of multiple linear regression. Fuzzy inference system proved that all factors affected the mode choice with a very strong correlation coefficient equal to 93.1% for general trips, 96.6% for work trips, 99.8% for shopping trips, 98.9% for education trips, 99.9% for recreation trips, and 99.7% for other trips. While in multiple linear regression models the most influential factors on the mode choice are car ownership, age and trip cost. The correlation coefficient equal to 27.8% for general trips, 31.9% for work trips, 20.4% for shopping trips, 27.9% for education trips, 21.1% for recreation trips, and 8.5% for other trips. Thus, it can be concluded that fuzzy logic models were more capable of capturing and integrating human knowledge in mode selection behavior.