Clustering Analysis Based On Consumer Behavior Data Of Electric Merchants

With the rapid development of internet technology,various products of the internet are bringing unlimited convenience to our lives.But with the advent of the information explosion,a large number of users' behavior daily data is collected.Facing these massive data,effective data mining work can help enterprises to dig deeper information and find new income-generating points for enterprises.On the other hand,it can help increasing customer viscosity by the way of personalized recommendation based on user historical consumer behavior data.Based on the consideration of increasing user viscosity,the paper makes use of the historical data of consumer behavior to cluster.In order to achieve personalized recommended role,this research analyses other information in the same group of consumers by mining the law of its consumption behavior.In general,clustering algorithm is used in the classification of crowd,which K-means clustering algorithm is widely used because of its simple logic and the understandable result.By comparing the distance between the sample points and the heart,the traditional K-means clustering take samples into different clusters,usually the Euclidean distance is used.However,Euclidean distance does not distinguish the importance of sample features.In other words,it does not take into account the differences among sample features,which results in inaccurate expression of information in clustering.In order to correct the deficiency of the K-means algorithm,this paper uses the concept of entropy in the information theory to modify the user's characteristic vector,so as to fully describe the user's behavior.In addition,the PSO optimization algorithm is adopted to optimize the initial heart selection.Finally,the improved K-means clustering results show that the algorithm in this scenario can significantly improve the performance of clustering algorithm,and the clustering effect of users is more significant.