A Study On The Fund Expenditure In The New Rural Cooperative Medical Service System Based On Independent Component Analysis

In our country countryside the New Rural Cooperative Medical Servicesystem(NRCMS) is basic medical security system，in which the fund management isthe key point and the difficulty. The NRCMS management office must carry on theregulation of the fund expenditure during the NRCMS operation in order that the fundis balanced on revenues and expenditures and it is used effectively andreasonably．The great uncertainty of the fund expenditures，caused by the complexoperation mechanism and the wide involving range of NRCMS and the numerousinfluencing factors，brings the very major difficulty for the regulation of the fund．Inorder to supervise and regulate the NRCMS fund more effectively，the author intendsto find the factors that influence the fund expenditures and how the factors influencethe fund expenditures through the analysis of the historical data that have beenaccumulated for years. And in order to provide the reference for the relationalpolicy-making and adjustment of the compensation mechanism, the data for fundexpenditures in coming time is made the forecast according to the historical data.In the signal processing field the independent component analysis(ICA) is a newblind source separation technique to seek for internal factors and components from themulti-dimensional statistical data. It can reveal the independent factors hidden in thecomplex phenomena, so it has been applied in financial field as an effective tool thatcan mine potential influencing factors. As a example, the expenditure data of thehospitalization compensation fund in20towns and offices in a city is studies by usingindependent component analysis method in this article．First，the FastICA algorithm is used to analyze the fund expenditure data of the20towns and offices．And four independent components are obtained．The firsrindependent component represents the influence of the actual hospitalization expenseto the fund expenditure．The second independent component represents the influenceof the compensation proportion to the the fund expenditure．The significance of thethird independent component is not explicit，so that it can’t been made the very goodexplanation．It may represent the influence of the medical behavior in hospitals to the fund expenditure．The fourth independent component represents the change of thefund expenditure caused by some factors such as the deductible and the ceiling ofcompensation．According to the hybrid matrix separated,the weight of the firstindependent component is the biggest and overwhelming in the four components．So,the fund management should focus on the control of the actual hospitalization expensebecause of the greatest impact of the actual hospitalization expense to the fundexpenditure． On the other hand,the smallest weight of the fourth independentcomponent indicates that the level in the adjustment and control based on furtherreducing the deductible is low in the case of very low deductible．Next,the prediction based on the independent component separated for theincoming fund expenditure is carried on by the support vector machine method．Theresearch result show that the prediction in independent component space is moreaccurate than in primary data space by the support vector machine method．The research in this article shows that the independent component analysismethod can been applied to mine the factors that influence the NRCMS fundexpenditure in order that how the NRCMS offce supervises and manages the fundmore effectively．Furthermore, the prediction based on the independent componentseparated for the incoming fund expenditure,by the support vector machine method,can play an important role in providing reference about how to make the relationalpolicy and adjust the compensation mechanism timely.