Research on Taxi Travel Time Prediction Based on GBDT Machine Learning Method

摘要

Predicting taxi travel times throughout the city can help organize taxi teams and minimize waiting times for passengers and drivers, creating smart cities. In this paper, we propose to use the Gradient Lift Decision Tree (GBDT) learning model to predict the future travel time based on historical taxi trajectory information. Remembering past information is very important here because future taxi requests are related to information about past actions that took place. For example, someone needs to go to a shopping center and predict the time to better organize friends and follow-up schedules. We proposed a multidimensional feature method based on machine learning gradient enhancement decision tree model. This method uses the original data to construct multi-dimensional basic features and inputs it into the GBDT model to obtain high-level features and further improve the prediction accuracy.