A collaborative filtering recommendation algorithm that simulates tf-idf similarity calculation

Abstract

The invention provides a novel method for calculating the similarity of users in the recommendation field. The method simulates a common term frequency-inverse document frequency (TF-IDF) statistical method of the search technology, the similarity of the users can be calculated without the need of repeatedly calculating a common scoring item value of a target user and a neighbor user and a particular scoring value of an object by the user, a prediction score of the object by the user is obtained through the obtained similarity, and finally a recommendation is made. The collaborative filtering recommendation algorithm simulating TF-IDF similarity calculation can effectively improve the accuracy and real-time performance of recommendation, and increases the expansibility of a recommendation system.

Description

technical field [0001] The invention belongs to the field of e-commerce recommendation systems, specifically relates to information retrieval technology and statistical methods, and realizes a brand-new similarity calculation method, which simulates the famous word frequency-inverse document frequency (TF-IDF) statistical method in information retrieval. A similarity calculation method that calculates the similarity between users or items in the recommendation algorithm to make recommendations. Background technique [0002] In recent years, with the rapid development of Internet technology, e-commerce has become a new fashion, forming a trend of rapid growth in recent years. E-commerce, which is a new business transaction process produced by the combination of IT technology and business behavior, is the main mode of business operation in the market economy in the 21st century. Through the e-commerce platform, people can enjoy the convenience of purchasing goods without leavi...

AI Technical Summary

Problems solved by technology

In traditional recommendation technology, methods such as cosine similarity and Pearson similarity are generally used to calculate user similarity, and these methods have a common feature, that is, they need to calculate the value of the item jointly rated by the target user and each neighbor one by one, and need to pass Specific scoring values ​​are used to calculate the similarity between users, so it takes a lot of time. Nowadays, most recommendation systems face the problem of "data overload", which challenges the real-time recommendation

With the increasing popularity of the recommendation system, more and more fields need to be applied to the recommendation system, and the traditional recommendation system cannot be well satisfied in other fields.

Images