using neural network for predicting hourly origin-destination matrices from trip data and environmental information

Predicting od demand has always been a challenging problem in transportation. conventional demand prediction methods mainly propose procedures for forecasting aggregated temporal origin-destination (od) flows. in other words, they are primarily unable to predict short-term demands. another limitation of these models is that they do not consider the impact of environmental conditions on trip patterns. furthermore, od demand prediction requires two individual steps of modeling: trip generation and trip distribution. this article presents a framework for predicting hourly od flows using the neural network. the proposed method utilizes trip patterns and environmental conditions for predicting demands in single-step modeling. a case study on new york city green taxi 2018 trip data is done to evaluate the method, and the results demonstrate that the network has reasonably accurate od flows predictions.