Year: 2019 Vol.: 68 No.: 2
Authors: Adrian Nicholas A. Corpuz and Joseph Ryan G. Lansangan
The Philippines is currently experiencing a housing backlog and is expected to reach 6.5 million by the year 2030 if nothing is done about this. It is in this context where the government and the private sector have partnered themselves to address the backlog. Financing institutions such as private banks and the Home Development Mutual Fund (i.e. Pag-IBIG) offer different home loans for Filipinos to be able to afford these houses. Using a local real estate development’s dataset, the study explores the application of predictive models in quickly determining whether a client will likely be able to get a home loan approved or not once he or she submits the preliminary documents for a home loan. Results show that in terms of accuracy, decision trees and random forest are superior in predicting home loan disapproval than binary logistic regression. The best predictive model is the random forest model, and results show that the main determinants of getting a home loan approved are loan equity term, total contract price of the house, equity payment status, and the income of the client.
Keywords: real estate, home loan, binomial logistic regression, decision tree, CART, CTree, CHAID, random forest