Agricultural material consumption data clustering method based on improved Chameleon algorithm

Abstract

The invention relates to an agricultural material consumption data clustering method based on an improved Chameleon algorithm. The method comprises the following steps: step one, obtaining a preference vector set X; step two, obtaining an initial weight graph and an initial module degree; step three, merging two nodes or two clusters corresponding to a minimum structure equivalent similarity so as to obtain a current weight graph and a current module degree; step four, if the current module degree is smaller than the initial module degree, outputting the initial weight graph as a clustering result, and otherwise, skipping to the fifth step; and step five, updating the current weight graph and the corresponding current module degree; and taking the current weight graph before updating as the initial weight graph, the current module degree of the current weight graph before the updating as the initial module degree and the current weight graph after the updating as the current weight graph, and skipping to the fourth step. The method reduces the application difficulty, enhances non-supervision of clustering, enables the clustering result to be more accurate and reduces deviations brought by manual arrangement of parameters.

Description

technical field [0001] The invention relates to a clustering method, in particular to a clustering method for agricultural material consumption data based on an improved Chameleon algorithm, and belongs to the technical field of agricultural material data clustering. Background technique [0002] With the perfection and maturity of e-commerce and the rapid development of agricultural industrialization, the e-commerce platform of agricultural materials products has gradually become an important channel for agricultural materials sales. Under normal circumstances, cluster mining of consumer behavior records on e-commerce platforms can help identify their preferences and purchase patterns, and lay a solid foundation for further personalized recommendations and behavior predictions. The current hierarchical clustering algorithms mainly include BIRCH algorithm, CURE algorithm, ROCK algorithm, Chameleon algorithm, etc. Among them, the Chameleon algorithm can find high-quality natu...

AI Technical Summary

Problems solved by technology

The current hierarchical clustering algorithms mainly include BIRCH algorithm, CURE algorithm, ROCK algorithm, Chameleon algorithm, etc. Among them, the Chameleon algorithm can find high-quality natural clusters of arbitrary shape, size and density and the characteristics of fast and efficient clustering algorithms. The algorithm has the following disadvantages: 1), the determination of the K value in the K-nearest neighbor graph needs to be done manually; 2), the selection of the minimum bisection is difficult; 3), the threshold of the similarity function needs to be given manually

These shortcomings increase the difficulty of use and affect the unsupervised nature of clustering

Images