Research On Rating Prediction Algorithm Based On Prospect Theory

The rapid development of modern network information has brought about the crazy growth of information.It is difficult for current consumers to find products that match their own from huge product information.The recommendation system was proposed for this problem,and many scholars considered different factors to improve the recommendation system.And the traditional matrix decomposition algorithm is proposed to extract the implicit relationship between goods and users from the interaction between users and commodities.However,the traditional matrix decomposition technique also has congenital shortcomings.The matrix-based decomposition-based scoring prediction algorithm pays more attention to the implicit relationship between goods and users,while ignoring the influence of other factors on the score prediction.Based on the traditional SVD algorithm,this paper considers the foreground theory and adds the influence of commodity history scoring factors,historical scores and users' own preferences to improve the algorithm.Firstly,through the statistical analysis of the behavior data of the user's purchase of goods,in addition to the influence of the historical score of the product on the user's score,the influence law is described by the loss avoidance effect of the foreground theory.Secondly,it analyzes the number of commodity history scores on the user's decision-making purchase,and through regression analysis,fits the corresponding regular curve.Thirdly,the commodity historical scoring factor and the commodity historical scoring number are combined with the matrix-based decomposition model based on the bias term,and the influence of the two factors on the commodity recommendation is analyzed.In the end,this paper verifies that the accuracy of the rating prediction algorithm in the recommendation system is improved by the rating prediction algorithm considering the commodity history rating factor and the commodity history rating.