Predicting Mortgage-Backed Securities Prepayment Risk Using Machine Learning Models

Abstract

An option to prepay a portion of a mortgage debt before it matures is included. This alternative, known as mortgage prepayment, puts the bank that provided the home loan at risk since it prevents them from receiving future interest payments and complicates their refinancing options. Prepayment risk refers to the possibility that borrowers would pay off their mortgages earlier than anticipated, lowering income flows to MBS investors. Actual prepayment rates that are greater or lower than anticipated might reduce cash flows while increasing the risk of extension for investors., Predicting the mortgage-backed securities prepayment risk analysis is necessary for this project using the mortgage portfolio of “Freddie Mac” to predict mortgage borrower prepayment behavior. Formulating the prepayment analysis issue will be feasible using machine learning methods. Additionally, it looked at the distribution of the target variable and offered ideas for enhancing the regression model. The accuracy of ridge regression was achieved at 78%, then 89% accuracy for testing data using Logistic Regression, and, with the KNN model, achieved an accuracy of 76%. A user interface that asks for input from the user to enter the information and forecasts whether the customer’s mortgage will be paid off has also been created.