Improvement And Implementation Of Collaborative Filtering Recommendation Algorithm Based On The Combination Of Bundling Information

The research on recommendation systems has a long history,and it is widely used in e-commerce and other fields.Unlike other fields,the interactive information between users and commodities in the field of e-commerce is more abundant,and full use of this information can greatly improve the recommendation effect of the existing recommendation system.The bundling information refers to the mutual connection between the products established by users' behavior as the connection,such as being jointly purchased,commonly visited,and purchased after browsing,etc.Although these appear as the result of the user's subjective behavior,they actually reflect Objective links between commodities.This article mainly introduces the method of introducing and using product bundling information in collaborative filtering recommendation algorithm to improve the recommendation effect.This article focuses on three contents:The first is to use the product bundle information to fill the existing user rating matrix,design different weight distributions for different user behaviors,and selectively score the user's unrated items according to the relevance of the product bundle information.This article does research on product similarity,filling score design and so on.The second is to use the product bundle information to revise the existing user rating matrix.This article uses the inherent information such as the category and the label of the product to mine the connection between the product and the product and set the binding function.This connection does not depend on the user-product rating data.The improved user's rating of the product is composed of two parts: the first is the prediction of the rating of the item based on the characteristics of the user and the item itself,and the second is the rating of the item associated with the item to be rated to predict the item rating.The third is to use the product bundling information to improve the existing neural network recommendation algorithm.The algorithm uses user,product and product type information as neural network input,and improves the recommendation effect by increasing the input data dimension.The specific method is to embed user information and product information into a neural network to generate a user feature matrix and a product feature matrix.Product information needs to consider the influence of product categories.Finally,by performing a dot product operation on thecombined user feature vector and commodity feature vector,the predicted result can be obtained.The experiment uses the Musical Instruments data set obtained from the Amazon data set,and based on the common purchase relationship(relate attribute)of the four commodities in the data set,rationally design the implementation methods of the above three research contents.At the end of the article,the relevant experimental results are explained in detail,and a recommendation system with bundling information is added to verify the effectiveness of the bundling recommendation system.Comparing the results with the results obtained by the conventional collaborative filtering algorithm,the experiment shows that by increasing the score filling based on the product bundling information,changing the bundling coefficient,and increasing the input of the neural network recommendation algorithm,the accuracy of the user's product rating can be improved in a specific environment And improve the accuracy of the recommended list.