Research And Implementation Of Hybrid Recommendation System Combining Attribute Features

In the context of the rapid development of today's e-commerce platform,collaborative filtering algorithm is one of the most popular personalized recommendations in e-commerce,has reduced many valuable time for users when selecting products.However,there are still many problems that need to be solved and improved,such as the sparseness of data,cold start of new projects or new users,and insufficient coverage of recommendations.These problems also require researchers to continue to work hard to improve,the author of this paper has proposed his own improvement programs,the specific main work is:(1)A hybrid pre-processing filling scheme used in the case where the project attribute is used to calculate the similarity between items and the user's scoring characteristics is combined to calculate the predicted score,and the scoring data is more extremely sparse is proposed.As the size of the recommendation system grows,the system data will appear sparse.For this type of problem,this paper proposes a corresponding improvement plan.The traditional collaborative filtering algorithm only considers the user's scoring data when calculating the similarity between users or projects.There are many disadvantages in the way of relying on a certain data parameter to calculate the similarity.For example,when the score data is extremely lacking,the final similarity may be 0 due to too few items commonly scored among users in the process of calculating similarity;or when a new project has just been added to the system,the similarity to other projects cannot be calculated because the new project does not have any user ratings.(2)A collaborative filtering algorithm is proposed that considers both the user's attribute characteristics and the user's rating data for the project.The traditional collaborative filtering algorithm is very inappropriate in the process of calculating the similarity and selecting the nearest neighbors to obtain the nearest neighbors only by relying on the user's scoring data.This practice will not only lead to a decrease in the recommendation accuracy of the recommendation system,but also cause the cold start problem of the recommendation system.The improved algorithm proposed in this paper needs to consider not only the existing data such as user score data,but also more objective and stable data such as user attribute characteristics when selecting the nearest neighbor.Using the similarity calculated by the user attribute feature and the similarity calculated by the user's score data,the two obtain a final similarity by a certain weighting calculation method,and perform the nearest neighbor selection based on the similarity degree.On the basis of this similarity,the nearest neighbor selection is performed,and then the prediction score of the item is calculated by using the score data of the nearest neighbor,and the recommendation is given to the user by the ranking of the predicted score.(3)In this paper,the experimental data set is improved,so that it can meet the system design requirements,and the improved collaborative filtering algorithm is used to realize the design of the movie recommendation system based on the data set.