Predicting residential property value: a comparison of multiple regression techniques

Date

2022-10-24

Department

Program

Citation of Original Publication

Whieldon, L., Ashqar, H.I. Predicting residential property value: a comparison of multiple regression techniques. SN Bus Econ 2, 178 (2022). https://doi.org/10.1007/s43546-022-00358-4

Rights

This item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.

Subjects

Abstract

This study aims to predict the prices of residential properties in Catonsville, MD based on publicly available tax assessment data maintained by Maryland government institutions. Recent legislation in Maryland has made it a requirement for local governments to provide up-to-date, easily accessible data to its constituents. This study used three regression techniques: Linear, Ridge, and Lasso regression. Extracting over 11,000 property records from Maryland’s Open Data Portal (ODP), we applied regression techniques to predict the price. We used various independent features to predict the price of residential properties including prior year housing sales price, size of house, house age, street address type (single family or townhouse), if the house has a basement, dwelling type, number of stories, and dwelling grade (scale from 1 to 6). Outperforming Lasso and Ridge regression, Linear regression can enhance the predictability of housing prices and significantly contribute to the correct evaluation of real estate price using only two years of historical data and without additional socioeconomic variables. We also found that among others; prior year housing sales price (positive), size of house (negative), and house age (positive) are the most significant factors in predicting housing sales price in Catonsville, MD.