Abstract:
The abnormal pattern recognition of district water-using becomes the vital basis of water companies' scientific management.This paper would combine the density-based clustering algorithm with k-means algorithm,which ould do abnormal pattern recognition to urban water supply networks.Initially,this paper would extract time and instant water-using amount as features and all data samples were separated into different patterns by k-means algorithm.Then it built the density-based clustering algorithm to identify outliers of different water-using patterns.In this paper,the monitoring data of two districts in a certain place were tested,and the recognition results and analysis of abnormal water-using patterns were obtained.Compared with existing anomaly detection schemes,the proposed fusion algorithm in this paper had gotten more intact and accurate results.