Heat-supply Network State Prediction Based On Optimum Combination Model Of Data Mining

Nowdays, people pay more attention to operating the network more securely and stablywith the widely using in heating network management system of our country. The real-timemonitoring technology of heating network is not perfect, the huge database forms the datagrave. When facing all sorts of drawbacks, it is particularly important to take effective way toreuse the historical data and improve the existing systems.The first part introduced the basic theory of data pretreatment and entire process in thesummary, then recommended several classic methods of decision attribute discretization,compared the advantages and the disadvantages of each algorithm; finally chosen the decisionattribute discretization of rough set based on information entropy and attribute importancewhich pre-processed massive network data, to lay the foundation for the later stages.The session of Data mining mainly adopted the following kinds of mining algorithms:the prediction of classification based on the decision tree induction, cluster analysis based onthe K-means partition, and association rules mining based on frequent item sets model.Aiming at the defects of three classical algorithms: inductive process with deviation,randomly initial point and scanning the database repeatedly, we used different optimizationschemes to avoid defects, then conducted the simulation experiment, compared the efficiencyof the optimization algorithms and the classic algorithms, and verified the availability of theoptimization model, finally by using the optimization ID3algorithm, the K-means withcertain initial points, the distributed Apriori, we mined the final prediction rules which laid thefoundation for establishing combination forecasting model.According to the three kinds of mining model which were used in the predictioncombination model: decision tree classification, clustering analysis, frequent item set patterns,firstly, we evaluated the importance of every algorithms, and established the combinationalgorithm to determine the models’ weights, so that we could get the prediction formula ofcombination model, then embedded the created model into the database so as to forecast theheat-exchange station in the network, finally completed real-time monitoring and predictedaccurately.The topic provided a scientific and convenient way of real-time monitoring inheat-supply network combining the data mining and forecasting model. The result shows that:the implementation of the above algorithms had some instructional function for theimplementing state prediction reasonably and using the system properly, this method providesa more reliable foundation for intelligent monitoring and heating network diagnosing.