An optimal clustering algorithm selection method and device based on multi-attribute decision

Abstract

The invention discloses an optimal clustering algorithm selection method and device based on multi-attribute decision, and the method comprises the following steps: employing a plurality of to-be-selected clustering algorithms, and carrying out the clustering of a data set; Evaluating the clustering result of each to-be-selected clustering algorithm by adopting a clustering effectiveness evaluation method, and constructing a decision matrix on the basis of the evaluation result; Calculating a weight value of each column of the decision matrix; Processing the decision matrix by adopting an MSMrelated operator to obtain a final evaluation value of each to-be-selected clustering algorithm; And selecting an optimal clustering algorithm from the plurality of to-be-selected clustering algorithms according to the final evaluation value of each to-be-selected clustering algorithm. According to the method, a clustering algorithm selection framework is established, so that the problem that an experience-based clustering algorithm may not obtain a good clustering result can be effectively solved, several different evaluation methods are selected and combined with a processing result of the evaluation methods to verify the clustering result, and the selected algorithm is more accurate.

Description

technical field [0001] The disclosure relates to a method and device for selecting an optimal clustering algorithm based on multi-attribute decision-making. Background technique [0002] With the rapid development of society, many industries have produced a large amount of data, and various information technologies such as artificial intelligence and data mining have been applied in many aspects. Clustering is a commonly used data mining method to identify potentially associated distributions and patterns in data. Clustering is an unsupervised data processing method without prior data, so it depends entirely on the similarity between data. Due to the unsupervised nature of clustering, how to measure the performance and correctness of an algorithm is critical. Furthermore, the no free lunch theorem ever states that the universality of an optimal method or model does not exist. Some clustering algorithms may work for convex structures but not for ring cluster structures, su...

AI Technical Summary

Problems solved by technology

[0005] In order to overcome the deficiencies of the prior art above, the present disclosure provides a method and device for optimal clustering algorithm selection based on multi-attribute decision-making. Establishing a clustering algorithm selection framework can effectively solve the problem that empirical clustering algorithms may not be able to obtain good results. For the problem of clustering results, several different evaluation methods were selected and combined with their processing results to verify the clustering results, making the selected algorithm more accurate

Images