Fee-paying behavior analysis method based on SOM neural network clustering algorithm

Abstract

The invention discloses a fee-paying behavior analysis method based on the SOM neural network clustering algorithm. The method comprises steps that data of all the fee-paying user basic attribute information and the fee-paying habit attribute information of an entire region is acquired to form a data set; behavior index parameters, the customer classification quantity and connection right constraint conditions of the data set are determined, and the SOM neural network is constructed; a part of samples in the data set is selected, training of each learning mode of the SOM neural network is sequentially carried out, and each of the connection weights connected with the winning neuron is continuously optimized and corrected until the correction amount satisfies the set value; the data set isclassified through utilizing the optimized SOM neural network to acquire target behavior index parameters, the classification result of the customer classification quantity is satisfied, an average value of each index of the data set is calculated, and the fee-paying behavior clustering result is acquired. The method is advantaged in that certain relevance among influence factors can be acquired through data mining, and classification and further research of client fee-paying behaviors are further facilitated.

Description

technical field [0001] The invention relates to a payment behavior analysis method based on a SOM neural network clustering algorithm. Background technique [0002] With the development of diversification of payment channels and payment methods, the original "single payment" mode of electric power agency business office payment has been broken. The preferred method is that individual business offices are overcrowded during peak payment periods. The contradiction between users' payment habits, payment needs and payment channel construction has become prominent, and problems such as hidden dangers of electricity fee recovery, hidden dangers of power supply services, and unreasonable allocation of human resources have gradually emerged. [0003] The application of electric power big data is not only technological progress, but also involves major changes in the development concept, management system and technical route of the entire power system in the era of big data. It is t...

AI Technical Summary

Problems solved by technology

[0006] At present, the commonly used clustering methods mostly use the K-means algorithm, but it has the problem that it is greatly affected by abnormal factors. In the technical field of electricity payment, where there are some abnormal customers, the application effect is not good, and it cannot be very good. Get the correlation between various user data

Images