Mixed attribute feature large data set clustering algorithm based on CSA

Abstract

The invention discloses a mixed attribute feature large data set clustering algorithm based on a CSA. The algorithm comprises the following steps: S1, an antibody community is initialized; S2, clone operation is carried out; S3, immunization gene operation is carried out; S4, clone selection operation is carried out; S5, clone death operation is carried out; S6, single-step operator iteration is carried out; and S7, antibody encoding is carried out, and steps from the second to the sixth are repeated until a clustering archetype is converged to an optimal solution. The clonal selection algorithm (CSA) is used for large data set clustering analysis. The clustering performance is evaluated in the large data set, and an experiment result shows that the method can effectively find out the clustering structure in the data. When clustering analysis is carried out on a large amount of data with numerical values and generic mixed features, the algorithm based on the CSA is quick in convergence speed and does not depend on selection of the initial archetype, and convergence to the global optimal solution with the possibility of 1 can be realized.

Description

technical field [0001] The present invention relates to a large data set clustering algorithm, in particular to a CSA-based mixed attribute feature large data set clustering algorithm. Background technique [0002] In the prior art, in data mining, we often encounter and analyze a large amount of data with numerical and categorical characteristics. However, most of the existing classification algorithms can only deal with numerical feature data or categorical feature data alone, but cannot analyze data with two mixed attributes. [0003] Dividing a sample set into various classes is a basic operation in data mining, and has been widely used in many tasks, such as classification (unsupervised), aggregation, division or dissection, etc. Clustering is a very popular approximate partition method. It divides a group of samples into several categories, so that under a certain standard, the samples in the same category are close to each other, and the differences between samples ...

AI Technical Summary

Problems solved by technology

Similar to the k2 prototype algorithm, this new algorithm is also sensitive to prototype initialization and is easy to fall into local extreme points. Therefore, a clustering method based on genetic algorithm (GA, GeneticAlgo2rithm) was proposed, although this method can converge with a higher probability To the global optimal point, but the convergence speed is slow, and it is prone to premature phenomenon

Images