The Research On Financial Performance Prediction Using The Least Squares Support Vector Machine

Financial distress prediction and research has entered a relatively mature stage sinceAltman began poineering research on it. However the research might give the forcasters afeeling of too late because the standard: weather a company is special trated. Because ofthis, scholars put forward a reserch: prediction and research of financial performance. Theprediction and research of financial performance is still in an initial stage. It providesinformation required for stakeholders in decision-making by predicting the futureperformance of the company. The purpose of this study is to chose a performanceprediction method for manufacturing to provide information for the stakeholders.This article is divided into six parts: the first part describes the researchingbackground, researching meaning and references. The second part of the article studies thetheoretical basis, including the statistical learning theory and the ralated theory of supportvector machine. The third part describes the selection and pretreatment of the predictors.There are24indicators which reflect the company’s solvency, profitability, operationalcapabilities, the ability to grow and cash floew capacity. After taken normal distributiontest, significant difference test and factor analysis, two common factors was extracted. PartIV is the empirical research of the article. The paper chose642manufacturing listedcompanies in2010initially, after removed the ST companies as well as those whoseperformance increasing or decreasing less than50%, the final number of the treated sampleis264(of which there are132performance increase as well as132performance decrease),these264samples were divided into training set and a sample set. There are88trainingsamples (44results rise in the company,44decline in performance company) and44testsamples (22increase in performance companies as well as22decrease in theperformance).Logit regression model and the LS-SVM model were constructed and thepredicting results of them were compared. The fifth part is the case of LS-SVM modelusing. Part VI is the conclusion of the article as well as the shortcomings.The least squares support vector machine(LS-SVM) was introduced in forecastingprocess of financial results and Semi-annual data was used to predict. The result is that: the prediction accuracy of LS-SVM model is higher than that of the Logit regression model.The main innovation of this paper contains the following two sides.First, the least squares support vector machine was applied to the financialperformance forecast. Through the study we found that the prediction accuracy of theLS-SVM model is64.8%, which is slightly higher than that of the Logit regression model.Second, the article using semi-annual indicators for forecast. During the study,semi-annual data was used which might provide financial results forecast in time.