Personalized behavior recommendation method based on federal learning

Abstract

The invention discloses a personalized behavior recommendation method based on federal learning, and the method comprises the following steps: 1, collecting user behavior characteristic data, classifying the user behavior characteristic data, and dividing local clients according to the types of the user behavior characteristic data; standardizing the user behavior characteristic data, and constructing the standardized user behavior characteristic data into a characteristic matrix form; 2, establishing a local model at a local client; and 3, adding a dual subjective logic model in a federal learning algorithm based on interaction between the server side and the client side user, quantifying a reputation value of a local model, and finally uploading the trained model parameters and the reputation value to the server. Privacy security is facilitated, and the privacy of the user is stored in a local client; and group recommendation can also be carried out on the basis of recommendation of individual users.

Description

technical field [0001] The invention relates to the field of machine learning, in particular to a federated learning-based personalized behavior recommendation method. Background technique [0002] In recent years, with the development of Internet technology, society has entered an era of information explosion. Massive information appears at the same time. On the one hand, people's privacy is easily touched and violated. On the other hand, it is necessary to learn people's behavior habits, thus forming Targeted recommendation system. Therefore, federated learning is designed to analyze data without touching it. However, the existing federated learning algorithms are still lacking in the analysis of personalization, and this field is very commercially valuable. [0003] The patent application number is 202121739511.7, which discloses a flexible and combinable liquid crystal screen intelligent display system. It is an information display system with a central network structu...

AI Technical Summary

Problems solved by technology

[0004] Problem 1: The problem of protecting user privacy has not been resolved

However, the existing laws do not allow companies to obtain users’ private information at will, so it is necessary to consider completing personalized recommendations on the basis of protecting users’ privacy. Laws at home and abroad are now prohibiting companies and individuals from illegally using private data

[0005] Problem 2: The existing recommendation methods are mainly for individuals, not for an age group or a group. The efficiency of recommendation is relatively low, and group activities require uniform movements, clothing or clothing for a group of people. Slogans etc.

Images