Personalized Recommendation Based On Changes Of User Interests And Sorting

At present,due to the rapid development of network technology,the Internet data also followed the increase dramatically.Thus people come into an information era from the block to an information era of big data.In the era of big data,people can not find the information they need from large amounts of data quickly,which reduces the efficiency of user access to information.This is the general problem of "information overload".An effective solution to the problem of information overload is the recommendation system.The recommendation system can spontaneously find useful information for users.Now the recommendation system research now gets the favour of many scholars.To a large extent,some success has been achieved.However,with the increasing amount of data in recommendation system and the complexity of recommendation requirements,the problems in recommendation system are also gradually highlighted.Aimed at the problems and challenges facing the recommendation system,this paper proposes a recommendation model based on the change of user interest and personality sort.The model is divided into two modules including score prediction module based on user interest change and initial recommendation list reordering module based on sort.The former mainly solves the inaccuracy problem of user interest changes,resulting in an initial recommendation list.The latter mainly solves the sorting problem of the recommended results,and finally several items in the front are recommended to the user.In the score prediction module,the paper proposes the algorithm based on the change of user interest.Firstly,the algorithm establishes time utility function according to the Ebbinghaus forgetting curve.Then the time-weighted k-means clustering is performed on the project.and finally through the calculation of time-weighted similarity,the algorithm can find the nearest neighbor,so as to generate the initial recommendation list.In the initial stage of the sort of recommended list this paper puts forward the reordering algorithm.Firstly the algorithm caryys on feature construction of the score data,then determines the training samples.And finally the algorithm generates sorting Learning model based point-wise.The algorithm can input the initial recommendation list generated by the score prediction module into the model,calculate the score of the items in the list,and then sort.Finally,the items in the front are recommended to the user.Through experiment,this paper validates the effectiveness of the proposed recommendation algorithm based on user interest changes and personality sort.The experiment shows that the proposed algorithm is outstanding on many indexes such as MAE,RMSE,Precision and MAP.So the proposed algorithm is feasible.