Addressing Societal Challenges Through Analytics: A Framework for Building a Foreclosure Prediction Model Using Publicly-Available Demographic Data, GIS, and Machine Learning

Abstract

Information systems (IS) and data analytics-focused academic disciplines remained surprisingly silent in attempting to contribute to a public understanding of critical societal challenges such as foreclosures. This paper tackles the gap by presenting a framework for building foreclosure prediction models by integrating publicly-available census-tract demographic data and readily-available technology (geographic IS (GIS) and machine learning (ML)). The framework is tested and validated using over 19,000 foreclosures from Cuyahoga County (OH) using J48 decision tree, artificial neural network, and Naive Bayes algorithms. The framework’s empirical test identifies nine critical demographic attributes to successfully predict foreclosures, confirming the findings of prior studies while offering several new, highly predictive variables that were missed by prior research. This research is a call to broader IS, CS, and data science communities to assist society in understanding critical societal issues that may need deploying and integrating more advanced technologies.