A Movie Hybrid Recommendation Algorithm Integrating Bias And Auxiliary Information

Facing the huge amount of information,the recommendation system is one of the effective ways to solve the problem of information overload.The traditional collaborative filtering recommendation algorithm mainly uses scoring for prediction and recommendation.The method is simple and effective,but the disadvantages of the collaborative filtering recommendation algorithm are also obvious.First,collaborative filtering simply performs linear interaction between the user and the project to predict the score,ignoring the non-linear relationship,which makes the recommendation accuracy limited.Second,collaborative filtering only uses the rating information for recommendation,but due to the rating The sparseness of data makes it face data sparseness and cold start problems.Deep learning has developed faster and faster in recent years,and has been widely used in image recognition,natural language processing,and other fields.It has also gradually played an important role in the field of personalized recommendation.In view of this,this paper introduces neural networks and deep learning into recommendation algorithms,and combines the advantages of neural network's non-linear fitting and feature extraction to propose a hybrid recommendation model—AI-NeuMF + model.This article first introduced the background and significance of combining neural networks with recommendation algorithms and conducting hybrid recommendation research,and explained the basic theory and related technologies needed to build models.Secondly,based on the neural collaborative filtering model proposed by He Xiangnan and others,global offsets and time offsets are added on this basis.Global offsets take into account objective factors independent of users or items,and time offsets take into account user interests over time.The change.The model added with the bias term fully considers factors other than the scoring interaction to improve the accuracy of the prediction.Based on the recommendation based on scores,the attribute information and text information are incorporated to obtain a hybrid recommendation model—AI-NeuMF + model.This paper uses deep learning to extract features for auxiliary information,makes full use of the advantages of convolutional neural networks in natural language processing,and at the same time conducts multi-layer fully connected training on attribute information,so that attribute information reflects user or item characteristics more deeply.Finally,according to the model principle,the AI-NeuMF + model is used to verify the Movielens 1M data set,and the horizontal and vertical comparison with the classic algorithm is performed.This algorithm alleviates the problems of data sparseness and cold start to a certain extent,and improves the accuracy of recommendation.The AI-NeuMF + model combines the linear interaction advantages of matrix decomposition and the non-linear interaction advantages of multi-layer perceptrons,while adding auxiliary information combines the advantages of collaborative filtering recommendation and content-based recommendation algorithms,and adds global bias and time to the matrix decomposition module.Biasing further excavates the static and dynamic features of users and projects.In addition,the AI-NeuMF + model is used to verify real movie data sets.This algorithm has a certain improvement in recommendation performance,and has certain theoretical and practical significance.