Research On Stacking Fusion Model Recom-mendation Algorithm Based On Implicit Fee-back Feature ——Take Music And Reading Platforms As Examples

With the rapid development of science and technology,the ”Internet+” has been more and more widely used in various fields,such as digital music.The recommendation system can not only help users to quickly locate goals,but also help the company to maximize benefits.Therefore,the recommendation engine is popular,and its core part is so-called recommendation algorithms.Although traditional recommendation algorithms can solve information-overloaded problems to a certain extent,it is inefficient to solve the problems of data sparsity and insufficient personalization ability.Aiming at the above problems,this paper proposes a more efficient algorithm:stacking fusion model with implicit feedback feature.First of all,in order to improve the feature system and solve the problem of feature sparsity,based on the data of user interaction behavior,the algorithm constructs and extracts the feature by using multi-dimensional extraction method and time correction factor.Secondly,the stacking fusion model proposed in this paper is used to process the constructed feature to improve the accuracy of prediction.The stacking fusion model in this paper mainly takes the trained Light GBM,XGBoost and Cat Boost models as initial filters,and takes the predictied results of the three models as the input data of the logistic regression model to get the final predictied results.With the improvement of people's quality of life,music and reading have gradually become an indispensable part of people's life,and many music platforms and reading platforms have also emerge in large numbers.This paper makes an empirical analysis of music data set and news data set,and measures the effectiveness of stacking fusion recommendation algorithm through performance evaluation indicators.Experimental results show,compared with Light GBM algorithm,the multi-feature stacking fusion recommendation algorithm improves the AUC value by 17%,reaches 86% on the music data set;On the news data set,the AUC of the multi-feature stacking fusion reaches 82%.The values prove the fusion recommendation algorithm is effective.