D-010
User's Next Place Prediction based on User's Attributes and Significant Points Extracted from Trajectories

◎Xiasi Liu・菅沼　睦・亀山　渉（早大）

The aim of this study is to predict where people go next based on the current position, time and other context information. Using Microsoft GeoLife GPS Trajectories Dataset, by applying HDBSCAN to destination points and stay segments of all the trajectories, lists of significant places are firstly extracted. Then, with user’s attributes including time, weather, moving speed, the next place is predicted by applying machine learning algorithms. The experiment results show through these features high accuracy can be achieved, where some attributes, day of week, time of day, wind speed and temperature, are mainly contributed.