Research On Machine Learning To Risk Early Warning Model For State-owned Assets Supervision System

State-owned enterprises are the pillar industry of our national economy; thereforethe State-owned Assets Supervision problem becomes a concern of our government.Especially in the aspect of the timeliness and accuracy of the assets risk early warning isthe priority among the priorities to preserve and increase the state-owned assets valueand to prevent the loss of state-owned economy. The root of the problem of economicinformation delay for state-owned enterprises lies in the timeliness of asset risk warning.Because of too many human controlling factors in risk early warning scheme ofstate-owned enterprises operating performance of SASAC(State-owned AssetsSupervision and Administration Commission) comprehensive performance evaluationmeasures, it will be improved. Along with the expanding of information technologyapplication field, it is imperative now to make use of the modern intelligent technologyto construct the real-time supervision system of state-owned assets, and it is the futuredevelopment direction of the supervision of state-owned enterprises. The risk earlywarning model of real-time supervision system of state-owned assets requires theanalysis transfer from static analysis to dynamic analysis, from single variable judgmentmodel to the multivariable decision model, to make state-owned assets risk warningpossesses with prospective and reliability.This Paper is mainly study on the performance risk early warning of state-ownedassets. The main content structure about the Paper is as follows:In the first place, this paper makes a brief summary of the research by the domesticand foreign scholars in enterprise assets risk early warning field; introduces somecommon methods adopted by the domestic and foreign in the study on enterprises assetsrisk, as well as the advantages and disadvantages of its early warning system, then designthe general structure of this paper.In the second place, it summarizes the machinelearning methods; focuses on the theory method of the rough set attribute reduction,theory and model of time-series analysis, theory and algorithm of neural network, and thedefinition of wavelet theory.Thirdly, this paper mainly introduces the neural network model based on rough set,make classification against to the present business performance of the state-ownedenterprises. It takes the performance indicators data of state-owned assets in the CSMAR as test samples, and preprocesses the importing sample, then to use conditionalinformation entropy which is one of the attribute reduction methods conductingindicators reduction. The defect of BP algorithm is easy to fall into local minimum points,otherwise using the wavelet theory improved BP algorithm to do simulated experimenton data sample can get better convergence property and accuracy rate. Conduct waveletneural network training to the attribute reduction of sample data set and the original dataset to get the online learning results, compare these results to get the best modelparameters. The experiment result shows that the reduction state-owned assets index setis a good way to reflect the financial risk of state-owned enterprises. Based on the resultsof the indicators of attribute reduction, we do predictive modeling to future businessperformance risk of state-owned enterprises; mainly adopt the time-series analysismethod. Extracting the sample index sequence from the attributes reduction database,and then use time-series analyses model to predict the future trend of the state-ownedassets performance indicators; use the wavelet neural network (WNN) to do traininganalysis to the predictive values to get the risk early warning value. Do wavelet neuralnetwork training to the original samples sequence, then compare the experimentalsimulation results. The result shows that the reduction business indicators of state-ownedenterprises through time-series analysis prediction are able to accurately do the risk earlywarning.At last, the time series analysis model algorithm in the paper is applied to theintelligent information technology risk supervision system of state-owned assets. Thismakes a simply introduction to the design of the system structure and technical methodto realize the system; shows the option interface after the realization of risk early warningmodular.