Research On Hybrid Recommendation Algorithm Based On Matrix Filling And XGBoost

In recent years,the development of information technology has had a huge impact on people’s lives.In today’s society,the Internet has surpassed traditional media to become the main way for people to obtain information.However,the information data contained in the Internet is very large,and it takes a lot of time and energy to find the content that users are really interested in.The recommendation system can accurately recommend the required content for the user according to the user’s preference,saving the time and effort of searching and screening,so it is widely used in the information technology industry.The hybrid recommendation algorithm can overcome the shortcomings of a single algorithm.However,under the influence of data sparsity,the recommendation effect of this algorithm is often limited.Therefore,this paper studies the hybrid recommendation algorithm in the data sparse environment,and the detailed work is as follows:First of all,the theoretical knowledge related to the traditional recommendation algorithm is studied,and an improved Tanimoto similarity calculation method is proposed to solve the problem that the traditional Tanimoto similarity calculation method only considers the number of items rated by users and does not consider the difference in rating values.This method fully considers the size of the user’s rating value,and judges whether different users like the same item by setting a threshold function,and then can calculate the similarity between users more reasonably;the experiment shows that the improved similarity calculation formula can find the neighbor users of the target user more reasonably.The algorithm first uses the improved Tanimoto similarity calculation method to calculate the similarity between users,and combines the proposed prediction formula to prefill the missing values of the scoring matrix then use the matrix decomposition technique to further fill the scoring matrix.Experiments show that the performance of this algorithm is better than the traditional collaborative filtering recommendation algorithm based on the neighbor and the collaborative filtering recommendation algorithm based on the model,which can effectively improve the recommendation quality.Finally,aiming at the shortcomings of a single recommendation algorithm,a hybrid recommendation algorithm based on matrix filling and XGBoost is proposed.The algorithm first completes the sparse scoring matrix,and then uses the clustering method to cluster the users and items respectively,then calculates the membership degree of each user relative to each type of user and the degree of membership of each item relative to each type of item through the similarity calculation method.Finally,the features of users and items are constructed separately to form a data set suitable for supervised training,and the XGBoost model is used to train the data.The experimental results show that the hybrid recommendation algorithm based on matrix filling and XGBoost proposed in this paper has smaller prediction error than the previous algorithm,and can improve the quality of recommendation.