Power user clustering method of improved k-means algorithm based on genetic algorithm

The invention discloses a power user clustering method of an improved K-means algorithm based on genetic algorithm. The method comprises the steps that data is preprocessed and an initial group is generated; the difference degree of the data is calculated, and random k points are selected as a clustering center and are clustered; the difference degree of each cluster is calculated; the coordinatesof each cluster center are taken respectively, the number of coordinate points of the mean of the clustering points is taken as a characteristic, the characteristics of K clustering results are subjected to crossover and mutation, the obtained new sub-generation serves as an index of candidate clustering results; the fitness of the clustering results is calculated; the probability pi of being selected for each individual is calculated by the clustering fitness, selected individuals Xi for this round are filtered, and a queue Yi is generated; a new group X is formed by conducting crossover operation on the queue Yi, t