Recommendation model of deep neural network based on user rating preference correction
Aiming at the problem that users'product scoring habits vary greatly in the upstream and downstream product selection of the industrial chain,a User Preference score-correction based Deep Neural Network(UPDNN)was proposed based on user rating habits.Through the historical data,the proposed method first learned the rating preferences of each user,and a unique satisfaction projection function was designed to map user ratings into the sat-isfaction space for correction.Then,the recommendation model was trained and the satisfaction prediction of the product to be tested was performed by deep neural network in the satisfaction space,and eventually the user's Top-k recommended product table was given to achieve product recommendation.The experimental results showed that the recommendation results of UPDNN were more suitable for users'preferences than the classical recommendation algorithms on the Movielens dataset,which verified the effectiveness of the proposed method.
rating value correctiondeep neural networkinformation extractionfeature processing
万方数据
维普
