User behavior log density peak clustering method capable of automatically determining clustering center

Abstract

The invention discloses a user log density peak clustering method capable of automatically determining a clustering center. The user log density peak clustering method comprises the following steps: step 1, reading a data set in a user log data file; step 2, calculating the similarity between the two users; step 3, calculating an initialization truncation distance dc; step 4, calculating the density Rhoi of the user i, wherein the density value reflects the similarity index of each user; step 5, calculating the distance Deltai of the user point i; step 6, calculating the normalized Rho* and Delta* of the user i; step 7, searching data points with relatively high density; step 8, finding out data points with large distances; step 9, finding out a correct category center point; step 10, removing outlier users; step 11, classifying the behaviors of the users according to the found central points, and classifying the non-central points into categories which are mainly the main central users; and step 12, outputting a clustering result. The user log density peak clustering method has the advantages that the influence of subjective factors on the clustering effect is reduced, and the accuracy is high.

Description

technical field [0001] The invention relates to a method for clustering website log data, in particular to a method for automatically determining a cluster center and performing density peak clustering on website user behavior log data. Background technique [0002] In today's era of rapid information development, a large amount of log data of users browsing websites is generated every day, recommending users' preferences, whether the client's service needs to be improved, and what kind of behavior users generally have when browsing the web Habits, finding out the rules, improving service module settings, and making recommendations based on user preferences have become hot spots that all information portals need to pay attention to. Aiming at how to classify users with the same preferences and similar behavior trajectories into one category, dig out potential connections among them, and cluster users into different groups, this invention proposes an algorithm that can automa...

AI Technical Summary

Problems solved by technology

Aiming at how to classify users with the same preferences and similar behavior trajectories into one category, dig out potential connections among them, and cluster users into different groups, this invention proposes an algorithm that can automatically determine the cluster center, which can not only Browsing log data divides unlabeled users into different groups, and can also solve the problems of unbalanced user division, long division process, and large resource consumption

[0003] Now in the era of data explosion, traditional clustering analysis algorithms can no longer meet the requirements. In the process of data mining, how to cluster data quickly, efficiently and with low consumption has become the focus of many scholars.

Images