A Research On Outliers Mining Algorithm Based On Heat Metering Data

With the promotion of central heating and household-based heat metering in China,the internet of heat meters is becoming larger and larger. Peoples will not pay by area butthe heat measuring, equal cost for equal comfort amenity. The promotion of this systemshows equality of using heat in some way, however there are also some problems occurredduring the using of numbers of heat meters, such as breakdown of meters, stealing of heat,IC card cheat, etc. What’s more, the expansion of the internet of heat meters also results inaccumulation of historical data and temporary data in system database. Therefore, it issignificant to find some unusual situation in heat meters by data mining technology.Firstly, according to the theme of outliers mining based on heat metering data, theframe of remote heat meters data collecting system based on the WSN and GPRS isproposed. And designs the communication protocol of WSN data collecting module,communication instruction of GPRS data transmission modules and analytical format ofremote data packets. Then using SQL Server2005, describe the design of data Warehouseof heat meter information and the process of ETL.Secondly, the core R-HOT algorithm to high dimensional data of heat metering isproposed. This algorithm includes three stages: The first is to excavate strong associationrules with the frequent itemset mining algorithm based on rough hypergraph, and regardsheat meter record as vertexs in data warehouse, regards strong association rules ashyperedge, and calculate the weight of hyperedge by the given formula, then formheat-supply network hypergraph. The second is to divide hypergraph with GRASPalgorithm. The third is to test outliers with the given concept of outliers.Finally, combined the model of data warehouse on heat meter information, thefeasibility of R-HOT by testing data collected by the system of remote heat meters chargeand govern is proved.