Research On Multi-criteria Rating Recommendation Method Based On Time Effect

The emergence and spread of the Internet has brought a lot of information to users,and it has also caused difficulties for many users to obtain information that is really useful to them.The recommender system is considered to be an effective tool to alleviate the "information overload" problem.Since the multi-criteria rating has more user-specific preference information than a single overall score,incorporating multicriteria rating information into the recommendation decision process helps improve the quality of the recommendation,but the traditional multi-criteria recommender system study does not consider the time.The influence of factors on the recommended performance cannot meet the situation where the user's interest and project popularity change over time.Based on the above problems in multi-criteria scoring recommendation,this paper studies the multi-criteria recommendation method based on time effects based on the review of the research status at home and abroad.First,this paper integrates time information into a heuristic multi-criteria collaborative filtering algorithm to study the influence of user preferences over time and time information on recommended predictions.In the multi-criteria heuristic algorithm,the Ebbinghaus forgetting curve based on the forgetting rule is used to simulate the change of user interest,and a time penalty factor is introduced into the formula of the project similarity calculation to correct the similarity calculation result and track the popularity of the project.Then,this paper adds time window and criterion entropy to the EM-ANFIS model with good application effect on the multi-criteria scoring recommendation problem,pays attention to recent effective data,reduces the interference caused by multidimensional data,reduces the computational complexity,and improves the dynamics of the model.Finally,the two proposed algorithms are validated and analyzed on the dataset from the Yahoo movie website.It is found that the time-based multi-criteria collaborative filtering recommendation algorithm has higher accuracy than the traditional multi-criteria scoring recommendation algorithm.The time-based EMANFIS multi-criteria scoring recommendation algorithm extracts approximately 20% of the user's rating data in the near future.The accuracy of the recommendation and the recall rate are improved compared to the static model-based multi-criteria scoring recommendation algorithm.