School trip production modeling using an improved adaptivenetwork-based fuzzy inference system

Trip production has long been considered as a major element in trip demand estimation. Many models have been presented for this purpose. Models use socio-economic variables in order to predict trip production. This paper develops an adaptive-network-based fuzzy inference system (ANFIS) models to predict school trip production. ANFIS can construct an input-output mapping based on both human knowledge and stipulated input-output data pairs. In order to improve models' generalization capability, a heuristic algorithm is used to generate reasonable initial values for data loss in training data set. Models with different membership functions (MFs) were trained, validated and tested with real data obtained from Shiraz, a large city in Iran, and then compared with regression model made for school trip production. The results indicate that the improved ANFIS (IANFIS) with Gaussian MF performed more accurate than the conventional regression model