Classification and Prediction of Property Taxes using Various Machine Learning Models
Abstract
Accurate assessments of property taxes are important to inject more equity, transparency, and efficiency in local government revenue generation. Traditional statistical and rule-based approaches have often failed to consider the complexity of, and nonlinear relationships inherent with property features like location, area and market value which contributes to unreliability of tax assessment predictions. We leverage several machine learning (ML) models including Linear Regression, Decision Tree, Random Forest, Gradient Boosting, and Neural Networks to classify and predict property taxes using an extensive dataset of property features and historical tax assessments. We evaluate model performance using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> score. Experimental results show that ensemble learning methods including Gradient Boosting and Random Forest demonstrated improved accuracy and generalizability over classical statistical and deep learning modelling approaches. Our main conclusion is that machine learning shows the potential for a scalable alternative to property tax assessment automation. The implications of our study demonstrate a shift toward the modernized development of tax systems, adopting data-driven approaches for political formulation, fairness, and generating forecasting municipal revenue.