结合兴趣点类别周期属性和用户短期偏好特征的推荐模型OA北大核心CSTPCD

With the widespread application of location-based social networks in daily life,how to effectively extract users'hidden interests and behavior sequence patterns,and provide users with the next Point of Interest(POI)recommendation service to meet their personalized needs has become one of the hot issues in the recommendation field.Aiming at the problem of user preference mining in the next POI recommendation,this paper proposes a POI recommendation model CPSTIN(Combining Periodic and Spatio-Temporal Intervals'Network)based on the combination of periodic preference of user POI category and short-term interest.The model embeds the user record of signing in into the time window by hour period pattern,and uses the multi-head self-attention mechanism to extract the the user's periodic preference combined with the category of POI.At the same time,the model sends the discontinuous spatio-temporal interval information into the learnable matrix,and uses the linear interpolation method to extract the user's short-term interest based on high-order correlation.Finally,the validity of the model is verified on two real datasets.The model effectively uses the user's high-order relevance short-term interest and the periodic preference based on the POI category to more accurately predict the next POI that the user is most likely to visit.