POI Recommendation Based On Spatio-temporal Sensitive User's Long-term And Short-term Preferences

POI recommendation appears with the popularity of LBSNs,because the development of major software positioning technology can find every behavior track of users.In recent years,interest point recommendation has attracted much attention.By connecting users with interest points and pushing interest points that may be signed in for target users in the future,users can reduce the time to select interest points and better explore people's life rules.The point of interest recommendation generates a trajectory sequence according to the user's check-in history to mine the user's potential preference for the point of interest and recommend the POI that the user is most likely to sign in in in the future.However,due to the tall rareness of data,in the current POI recommendation ways,modeling through users' long-term and short-term preferences still makes insufficient use of temporal and spatial features,and some continuous POI recommendation methods can only predict where users may go next,but ignore when this behavior will occur,which will affect the accuracy of POI recommendation,Make the recommendation unreliable.Aiming at the above problems,this paper proposes a POI recommendation algorithm with time-space sensitive user long-term and short-term preferences.For the historical check-in trajectory data of user interest points,the user's short-term preference for interest points and the user's long-term preference for interest points are formed in turn,the user's final preference for interest points is obtained,and the user's possible check-in location in the future is predicted.The work of this paper is as follows:Firstly,preprocess the data set and form the user's sign in track in time mode according to the user's historical sign in record.Secondly,for the user's long-term preference,the user's check-in POI is encoded through LSTM.In view of the similarity of users' access to points of interest,the impact of time is considered.The daily check-in records of each user are divided into several different time windows,so as to capture the long-term preference of users for points of interest in a more fine-grained manner,and provide a time weighting operation for each time window,so that the model can make better use of the time mode.,Calculate the center distance for each user's daily track,and calculate the length between the center position of the historical sequence of interest points signed in by the user every day and the position of interest points signed in by the user last time.Under the influence of space,the long-term preference representation of users for POI is obtained.After that,the user's short-term preference for points of interest is related to the user's preference at the current time.The higher the user's preference for the nearest point of interest,the more likely the user is to check in at the location of this point of interest in the future.Finally,in order to verify the effectiveness of the model and the impact of important components of the model on the whole model,ablation comparison tests were carried out.The final results of the model show that the POI recommendation algorithm with time-space sensitive user long-term and short-term preferences can achieve better prediction results,and the model is more effective and accurate in the real data set.