Method for optimizing artificial fish school in clustering analysis mode

Abstract

The invention provides a method for optimizing an artificial fish school in a clustering analysis mode, and relates to the method for optimizing the artificial fish school. The method includes a first step of setting an original value of an artificial fish parameter, and calculating the food concentration of a current position by means of an objective function, a second step of executing a food searching module, a tailgating module, a clustering module and the like on each artificial fish according to an execution condition of a fish school module and advancing principles of artificial fishes, a third step of making the state of each artificial fish represent a decision variable and evaluating the optimizing degree of the fishes by means of calculation of the objective function, and a fourth step of choosing an input parameter of a K-center point according to bulletin board information and the position of the fish school and ultimately obtaining clear cluster division. According to the method, an artificial fish school algorithm is improved, optimization of data aggregation is carried out, needed time is reduced, and decision precision is high.

Description

technical field [0001] The invention relates to a method for optimizing an artificial fish school, in particular to a method for cluster analysis and optimization of an artificial fish school. Background technique [0002] At present, clustering research is mostly focused on hybrid clustering: although clustering analysis based on simulated annealing has good clustering results, this method takes too long to execute and is not suitable for the analysis of massive data. Although the hybrid clustering based on genetic algorithm has good parallelism, it still cannot solve the randomness of the initial cluster center, which leads to cluster aggregation and dispersion. Since the K center point method cannot automatically obtain the center point, it is necessary to compare the sum of dissimilarities each time to select the cluster center for the next iteration, so the computational cost of the algorithm will be very high. Contents of the invention [0003] The purpose of the pr...

AI Technical Summary

Problems solved by technology

[0002] At present, clustering research is mostly focused on hybrid clustering: although clustering analysis based on simulated annealing has good clustering results, this method takes too long to execute and is not suitable for the analysis of massive data

Although the hybrid clustering based on genetic algorithm has good parallelism, it still cannot solve the problem of random initial cluster center, which leads to cluster aggregation and dispersion.

Since the K center point method cannot automatically obtain the center point, it is necessary to compare the sum of dissimilarities each time to select the cluster center for the next iteration, so the calculation cost of the algorithm will be very high

Images