Broker Factors as Predictors of Mortgage Deal Success

Home Trust is a mid-size bank that provides innovative solutions to its customers. Home Trust provides mortgages for different property purchases including single-family dwellings, condominiums, duplexes, triplexes, store and apartment and rental properties. The process of funding a mortgage involves several interactions between Home Trust, brokers, underwriters, and the customer. We are committed to make sure that the mortgage process is as stress-free as possible for our customers and stakeholders. We evaluate each application individually to determine the appropriate pricing.

This project would focus on exploring factors (broker and client) and how they contribute to mortgage outcomes. Different data points would be explored as part of this project (for example: demographics, loan to value ratio, broker tenure and activity levels, brokerage size, average customer bureau scores, etc.). The goal is to understand how these different factors impact mortgage funding and predict the likelihood of mortgage deal success, utilizing machine learning to create a predictive model. Home Trust would leverage this predictive model to better understand our broker base, find efficiencies in our broker base to optimize our marketing opportunities as well as optimize service.

Interns working on this project would leverage SQL and Python to prepare the data, conduct relevant analyses and develop predictive model(s), such as Logistic Regression, XGBoost, Neural Network, Random Forest, and/or other relevant algorithms. The model(s) would be fine-tuned, evaluated, and presented, along with relevant findings and conclusions to Home Trust.

Faculty Supervisor:

Dmitry Krass

Student:

Partner:

Home Trust Co.

Discipline:

Business

Sector:

Finance and Insurance

University:

University of Toronto

Program: