A CNN and BP neural network-based recommendation method

Abstract

The invention discloses a CNN and BP neural network-based recommendation method, and the method mainly comprises the steps: estimating the hobbies of a target user through a series of behaviors of thetarget user, and recommending a possible favorite item sequence of the user to the target user. In the recommendation model, the core content is a model based on score prediction. Through the novel model CNN-BP provided by the invention, the next consumption behavior of the user is predicted, a convolutional neural network CNN is used for analyzing the behavior sequence, and the basic probabilityof each unwatched item is obtained; a final probability is calculated by using a back propagation neural network BP; And finally, the previous N items according to the final probability set is selected, and the previous N items is recommended as the items most likely to be watched by the target user next time to the user. According to the method, a movie recommendation model is taken as an example, and experimental results show that CNN- The BP model can predict the next viewing behavior of the target user more accurately, and provides more excellent performance for the recommendation system.

Description

technical field [0001] The invention relates to a recommendation model, and specifically provides a recommendation method based on CNN and BP neural network. Background technique [0002] In view of the various problems faced by the recommendation system, in the past various traditional recommendation algorithms, there have been many research results and widely used in various fields, such as content-based recommendation algorithms, knowledge-based recommendation algorithms and hybrid recommendation algorithms. etc. In recent years, deep learning has achieved good results in natural language processing and image processing, mainly including convolutional neural network CNN, recurrent neural network and BP neural network, and many scholars have applied neural networks to recommendation systems. [0003] In foreign countries, Falk J[1] and others proposed a recommendation system based on feed-forward neural network, and finally counted and evaluated the performance of the rec...

AI Technical Summary

Problems solved by technology

[0005] The calculation formula of the traditional prediction score has certain limitations, resulting in the accuracy of the recommendation is not very accurate, so this method proposes a new model to improve the accuracy of the recommendation

Images