Research On Context-aware Dynamic Personalized Poi Sequence Recommendation Method

With the development of mobile information technology,Location-Based Social Network(LBSN)has gradually become popular,and a large amount of check-in data has been derived.Abundant check-in data provide an opportunity to find out user behavior and realize personalized travel route recommendations.In order to save the time of people blindly searching for travel information,personalized travel route recommendations came into being.Personalized POI(Point Of Interest)sequence recommendation research is a key task of the personalized itinerary planning and personalized travel route recommendation,aiming to mine POI sequences that meet the personalized needs of users from a large amount of information.Generally,the external environment(e.g.,weather,time and so on)may change during travel,and the user's current check-in behavior is closely related to the external environment.Most of the existing researches are based on static environment for POI sequence recommendation,and it is difficult to recommend the POI that users like in real time according to the dynamically changing environment.Furthermore,people tend to visit popular POI subsequences.Therefore,this paper is dedicated to recommending the POI sequences that users are interested in and are popular with full consideration of weather and time changes.This paper first proposes a context-aware dynamic personalized POI sequence recommendation problem,and analyzes its complexity.Based on the user's check-in datasets,this paper analyzes the user's check-in behavior and the characteristics of POI sequence,and verifies the weather and time can affect the users' check-in behavior.Furthermore,there are certain rules in the migration behavior of users between POIs,and people are more inclined to visit some popular and classic POI subsequences.Then,this paper proposes a context-aware dynamic personalized POI sequence recommendation model(DPSR),which uses the Monte Carlo tree search(MCTS)algorithm to simulate the process of selecting POI in the dynamically changing environment based on the user interest and preference model with fine-grained contexts,the POI popularity model with fine-grained contexts,and the POI migration probabilistic model,and realizes the dynamic personalized POI sequence recommendation.In addition,the DPSR model also considers the different migration methods of users,and flexibly sets the migration speed,so that the recommended POI sequence is more in accord with the user's real travel behavior.Finally,this paper conducts some comparative experiments based on two LBSN checkin datasets,and uses 6 evaluation metrics,to evaluate its effect.Experimental results show that,the DPSR model can significantly improve the accuracy of the POI sequence recommendation system.What's more,DPSR can not only recommend POI sequences that meet the user's interest and preferences,but also ensure the popularity of the recommended POI sequences.