Swarm intelligence based behavior clustering system

Abstract

The present invention provides a swarm intelligence based behavior clustering system. The data representation of the system comprises a data structure and a data type, and a k-means mixed clustering algorithm is employed; the k-means mixed clustering algorithm which combines an ant colony clustering algorithm and a k-means clustering algorithm is employed and is mainly divided into two parts, a first part is used for carrying out ant colony clustering, a second part is used for collecting an ant colony clustering result by using a k-means algorithm; in the k-means mixed clustering algorithm, a similarity formula is similar to a fundamental model for ant colony clustering and an LF algorithm, but simpler probability transformation functions are adopted and are of two straight lines with the slope of k; and proved by subsequent experiments, due to improvements, both the precision and efficiency of the algorithm are more excellent compared with the existing algorithms.

Description

technical field [0001] The invention relates to the technical field of the Internet, in particular to a behavior clustering system based on swarm intelligence. Background technique [0002] At present, after the Internet industry has developed to a certain extent, the process of specialization and division of labor has also produced a more detailed differentiation in its internal structure, thus forming the entire Internet industry into several levels from low to high: Internet companies at different levels With different customer targets, service methods and profit sources, this naturally forms different business models of current Internet companies. To divide the structural level of the Internet industry from the perspective that is more common in the Internet industry at present, or to say that the business models of Internet companies mainly include the following types: [0003] (1) Access and Platforms: This is one of the initial business forms of Internet companies. ...

AI Technical Summary

Problems solved by technology

After sorting the users according to the size of the profit source of the website operator, we found that more than 80% of the profit of the website operator comes from 20% of the users (A), 70% of the users provide less than 20% of the profit (B), and the other 10% of users not only will not bring any benefits to the website operator, but will even weaken the profit level of the website operator (C)

The main defect of the two-factor combination method is that the user's dynamic description data is generally not considered in the classification process, and the user data is not fully utilized.

The shortcoming of the multi-factor combination method lies in the selection of factors that affect user classification.

Images