Simulation of Book Multi Feature Filtering Recommendation Targets under Deep Fusion of Interest
Due to the sparse nature of user rating behavior,which means that the interaction data between users and books is not rich enough,it is difficult to accurately explore users'interests and preferences,resulting in unsatis-factory book recommendation results.Therefore,a simulation method for book multi feature filtering recommendation considering deep interest fusion is proposed.Firstly,a progressive forgetting function was employed to calculate the in-terest feature weights.Then,the influence of the borrowing order and time on the feature word weights was used to ob-tain the weight of the book feature words.Secondly,the user's geographical location was identified.Moreover,the word vector was used as input data,and the emotional orientation was taken as output data.Furthermore,convolutional neu-ral networks were used to filter users'network check-in information,thus calculating the probability function of the in-terests and preferences published by users.Next,interests were integrated in depth.Meanwhile,the recommended item which was not in popular demand was calculated.Based on similarity,users were divided into corresponding clusters.Finally,the time-span factor was introduced,and then a collaborative filtering recommendation method was adopted to complete the goal of book multi-feature filtering recommendation.The simulation results show that the average absolute error value of the proposed method is around 0.2.And the user experience score is high.Therefore,this method can effectively improve the efficiency and accuracy of book filtering recommendations.
Deep fusion of interestsMulti-featureFilter recommendationsConvolutional neural networkWeight calculation
NETL
NSTL
万方数据
