The Prediction Model And Strategy Study Of Mortgage Prepayment

Mortgage prepayment breaks the normal cash flow planning of the bank, which cuts off the profit, improves the cost, and also stops the operation of the MBS .The mature prepayment prediction models of the developed countries are from empirical analysis, therefore our country hardly copy them because of the shortage of data and incompletely finance system.The developed countries pursuit the accurate prepayment prediction in solving the prepayment problem of MBS. However, because of the poor basic condition, this paper propose to eliminate the asset which will prepay in the stated period in the start of asset pool picking in order to reduce the uncertainty of the cash flow of MBS. According to construct the prepayment prediction model and add the additive contract terms to confirm the probability of prepayment in stated period ,which supplies the criterion of selecting the asset pool and provides the foundation of replying the prepayment for the bank .This paper picks up the prediction model and variable depending on the advanced literature, firstly weight on the variables by discriminant analysis and principal components analysis, and then cluster the samples with the two stage cluster. In the process of cluster, this paper concludes that the cluster based on discriminant analysis will make the better study effect when clustering to three groups. Afterward, this paper constructs the Logistic regression models and SVM models for the three groups, and also find the major factors in impacting the prepayment and the criterion of distinguishing the different groups. At meantime , this paper compares the Logistic regression model and SVM model , gets the conclusion that there is a better prediction accuracy in SVM, but the application is limited because of the worse explained characteristic. Then this paper refers to the two pieces of additive terms of the contract and designs the pricing the formula for the prepayment certificate, confirms whether to access to the MBS asset pool through competitive game. At last, this paper designs the prepayment prediction system to support the decision of the commercial bank by Visual Basic 6.0 and Access 2000.