Optimization of Genetic K-Means Algorithm Based on Greed

The results of k-means are easily influenced by the initial k-value and thus fall into a local optimum, and subsequent evolutions of the genetic k-means algorithm, although improved, can produce “premature” solutions. A greedy divide genetic algorithm is proposed to address this problem. By adding greedy ideas to the original genetic k-means, the algorithm optimizes the mutation operation in genetic inheritance so that it receives new elite offspring with a certain probability, thus improving the diversity in the whole population and avoiding the occurrence of premature maturity. To test the performance of the algorithm, it was tested on the Iris and Wine data sets, respectively, and the experiments showed that the GdGa-k-means method can effectively improve the classification accuracy and enhance the global merit-seeking ability.