Trip production modeling is the process of representing the effect of various socioeconomic parameters on human trip-making behavior. Whereas making a trip appears to be related to the sociocconomic characteristics through some functional form. This paper describes the use of an advanced type of NFSs, Adaptive Network-Based Fuzzy Inference Systems (ANFIS), for modeling trip production pattern. The proposed trip production model used a lour-step hybrid learning strategy. At first. by using a lincar regression model. a suitable initial situation was developed lar the NFS and then this initial system was trained until reach ing the final model. The data used in this research were collected from 55 traffic zones (47 zones in inner regions and 8 zones in outer regions) throughout Shiraz comprehensive study in 1990. In 1990's Sh iraz comprehensive study, linear regression analysis was used to model the trip production. In that study the models were developed for four major trips: work
trips, school trips, shopping trips, and recreational trips. In order to be comparable with the previous practice, we also used this classification and made the new models for those four trip purposes. We also use the same variables that were used in conventional models for building the new models. The predictions of the conventional models were compared with those from the new proposed models. The results indicate that the new models have capability to represent the relationship between the trip demands and the independent variables more accurately than the conventional models.
)