The Research On Recommendation Algorithm Based On Score Matrix

In the process of information recommendation, it is urgent that the problems of computing time and accuracy caused by the rapid data growth be solved.Improvement of the accuracy of the recommendation results and reduction of the error rate are of great significance in the recommendation system. The related study has been conducted in this thesis, and the main work includes following three aspects.Firstly, In this thesis, with the foundation of the experimental data, the characteristics of the user collaborative filtering recommendation algorithm(UserCF)and the Item collaborative filtering recommendation algorithm(ItemCF) are obtained,and according to the variation tendency of the user quantity and the item quantity of the score matrix data, experimental results are analyzed by using two above filtering algorithms. The results show that according to the feature of the score data, using the appropriate algorithm can make the algorithm more efficient, and achieve the goal of recommendation better.Secondly, In the environment of c + + programming language, with the score matrix being the data sources, the accuracy, recall rate, coverage and novelty value of UserCF and ItemCF under different K value are calculated, and the characteristics of UserCF and ItemCF are verified through experiments of four evaluation indicators.Based on the Characteristics of UserCF and ItemCF algorithm, this thesis analyzes the performance of recommendation algorithm suitable for different scenarios, and puts forward the recommendation algorithm based on the score matrix.Thirdly, The error rate and accuracy of the algorithm purposed in this thesis is determined by calculating the value of MAE. The thesis has completed the experiments based on MovieLens rating data and drawn the conclusion that determining the adopted algorithm based on the characteristics of the matrix and then recommending after calculating the similarity can achieve a better recommendation performance and that the time complexity degrades.