Multi-POI Recommendation Based On Genetic Algorithm And Ant Colony Algorithm

The emergence of positioning technology makes the personalized recommendation of POI(Point of Interest)as one of the key research areas in the field of recommendation.However,the current research fails to effectively discover the effect of spatial-temporal sequence of user check-in data on the recommendation,and the factors that recommend accuracy are also not taken into account.In this paper,user's check-in data based on location-based social network will be used to explore implicit spatial-temporal sequence features.The time factor is introduced into the genetic algorithm to encode the point of interest,predict the probability of crossover and mutation of different nodes,and finally combine the ant colony algorithm with the collaborative filtering algorithm to recommend point of interest for user.The main work of the dissertation is as follows:(1)The traditional point of interest recommendation algorithm mostly only considers the points visited by the user or the points visited by his friends,and seldom focuses on the sequence hidden information of the point of interest visited by the user.Therefore,use the check-in data in LBSN to build a point of interest transfer graph.Based on the point of interest transfer graph,explore the spatio-temporal sequence features existing in user's check-in data.(2)Introduce time factor into genetic algorithm.The user's visit to the point of interest is in a certain order.During a period of time,the point of interest visited by two different users are compared with the point of interest visited by other users in the current time period,and the more the times of visit are,the more likely they are to be accessed again,the more similar their preferences are to other users.Based on this,a fitness function is constructed and the search for point of interest using genetic algorithm increases the cumulative probability of distribution of the more active point of interest by an average of 7.8%.(3)When applying ant colony algorithm to search for point of interest,defines the user rating as pheromone,the pheromones also increase as the number of point of interest is increased.Combining the ant colony algorithm and the collaborative filtering algorithm,consider the user's interest changes,finally get the highest point of pheromone accumulation of interest,and according to the characteristics of space-time series of check-in data,recommended for the user POI sequence.(4)Finally,simulation experiments are carried out on the Foursquare dataset.The experiments show that the accuracy rate reaches 0.27 and the recall rate reaches 0.24.Compared with the current point-of-interest recommendation algorithm considering the time factor and the user rating,the research method proposed in this paper has greatly improved the accuracy rate and recall rate,and based on the research to build a point of interest recommendation system,provides a basic idea for building a point of interest recommendation system.