Machine Learning Approaches to Real Estate Market Prediction Problem: A Case Study

TL;DR

This paper develops and compares machine learning models, including XGBoost, for classifying whether property sale prices are above or below listing prices using a decade of real estate and socio-economic data.

Contribution

It introduces an integrated approach combining multiple machine learning algorithms with target encoding for property price classification.

Findings

XGBoost outperforms other models in accuracy and robustness.

The models achieve high precision and recall in classifying sale price categories.

The approach aids stakeholders in making informed real estate decisions.

Abstract

Home sale prices are formed given the transaction actors economic interests, which include government, real estate dealers, and the general public who buy or sell properties. Generating an accurate property price prediction model is a major challenge for the real estate market. This work develops a property price classification model using a ten year actual dataset, from January 2010 to November 2019. The real estate dataset is publicly available and was retrieved from Volusia County Property Appraiser of Florida website. In addition, socio-economic factors such as Gross Domestic Product, Consumer Price Index, Producer Price Index, House Price Index, and Effective Federal Funds Rate are collected and used in the prediction model. To solve this case study problem, several powerful machine learning algorithms, namely, Logistic Regression, Random Forest, Voting Classifier, and XGBoost, are employed. They are integrated with target encoding to develop an accurate property sale price prediction model with the aim to predict whether the closing sale price is greater than or less than the listing sale price. To assess the performance of the models, the accuracy, precision, recall, classification F1 score, and error rate of the models are determined. Among the four studied machine learning algorithms, XGBoost delivers superior results and robustness of the model compared to other models. The developed model can facilitate real estate investors, mortgage lenders and financial institutions to make better informed decisions.