Taxi single-region OD demand forecasting model based on Grid-LSTM

With the rapid growth of the Internet economy and the rising rate of urbanization, urban residents' travel modes have become increasingly diversified, with taxis featuring online booking becoming a common choice for daily commuting. Taxis meet the residents' "door-to-door" travel needs, but their travel modes exhibit significant randomness and uncertainty. Moreover, there are often imbalances in supply and demand, such as "difficult to hail a taxi" and "long-distance orders". Therefore, accurate and rapid prediction of taxi demand is crucial for improving regional transport capacity and achieving a "win-win" situation for platforms, drivers, and passengers. This paper uses the "New York City Taxi and Limousine Commission (TLC) Green Taxi Trip Data for April 2015" dataset, applying data mining techniques based on the LSTM taxi demand prediction model to forecast single-region taxi demand and regional OD matrix. The paper selects real network car order data to validate the models effectiveness. Additionally, different prediction models are compared to determine the optimal forecasting model.