Context-aware personalized POI sequence recommendation

Abstract

The Point Of Interest (POI) sequence recommendation applies to scenarios like itinerary and travel route planning which belongs to the class of NP-hard problem. What’s more, the external environment like the weather, time can affect the user’s check-in behavior such as people prefer to check-in in ice cream shop when the temperature is higher. We propose an algorithm to solve these problems that called Context-Aware Personalized POI sequence Recommendation based on reinforcement learning (CAPR for short). First, we model the users dynamic preferences that incorporate contextual information associated with the users’ sequence of check-ins. Then we use the Monte Carlo Tree Search algorithm to select the user-satisfied POI in different environments. What’s more, we can get the context-aware personalized POI sequence under the specified time limit. Finally, we test the proposed algorithm using an open source dataset. The experimental results show that the weather and time context can improve the accuracy of the recommendation. Our algorithm can improve the effectiveness of travel recommendations.