Applying Lasso Regression for Property Price Prediction in Emerging Markets: An Empirical Study of Prishtina's Real Estate Sector

Session

Management Business and Economy

Description

This study employs Lasso Regression (LR) to forecast apartment prices in Prishtina, analyzing a dataset of 1,468 transactions from the Kosovo Department of Property Taxes (2019-2023). Focusing on variables like distance to the central business district and apartment size, LR was selected for its ability to perform feature selection and regularization, applied using Python. The preliminary model included variables such as distance to central business district, distance from main road leading to central business district, size, floor, distance from bus stations, distance from ambulances, distance from schools, distance from green spaces, income status of inhabitants per apartment, and distance from public parking and after feature selection and regularization excluded variables such as the distance from the main road leading to the CBD, distance from bus stations, distance from ambulances and schools, and distance from public parking The methodology involved evaluating model accuracy with metrics such as Mean Squared Error and Coefficient of Determination. Emphasizing LR's feature selection, significant predictors were identified, streamlining the model and offering insights into Prishtina's property price determinants, highlighting LR's effectiveness in feature selection in the context of emerging markets real estate valuation. The Lasso Regression model applied to Prishtina's apartment prices demonstrated moderate predictive accuracy, with an R² of 0.501, indicating it captures about half of the market's variability. Performance metrics, including MAE, RMSE, and MSE, underscore the tradeoff between model simplicity and precision. Despite lacking a direct feature importance ranking, Lasso's feature selection capability identifies key variables, offering insights into Prishtina's real estate market. This study highlights the potential of Lasso Regression in feature selection and regularization within emerging markets, showcasing its strength in simplifying complex datasets for enhanced interpretability and actionable insights, thus marking a significant contribution to the field by suggesting to integrate Lasso regression with other ML models for better predictive accuracy.

Keywords:

Lasso Regression, Real Estate Valuation, Feature Selection, Feature regularization, Emerging Markets, Property Price Prediction

Proceedings Editor

Edmond Hajrizi

ISBN

978-9951-982-15-3

Location

UBT Kampus, Lipjan

Start Date

25-10-2024 9:00 AM

End Date

27-10-2024 6:00 PM

DOI

10.3107/ubt-ic.2024.29

This document is currently not available here.

Share

COinS
 
Oct 25th, 9:00 AM Oct 27th, 6:00 PM

Applying Lasso Regression for Property Price Prediction in Emerging Markets: An Empirical Study of Prishtina's Real Estate Sector

UBT Kampus, Lipjan

This study employs Lasso Regression (LR) to forecast apartment prices in Prishtina, analyzing a dataset of 1,468 transactions from the Kosovo Department of Property Taxes (2019-2023). Focusing on variables like distance to the central business district and apartment size, LR was selected for its ability to perform feature selection and regularization, applied using Python. The preliminary model included variables such as distance to central business district, distance from main road leading to central business district, size, floor, distance from bus stations, distance from ambulances, distance from schools, distance from green spaces, income status of inhabitants per apartment, and distance from public parking and after feature selection and regularization excluded variables such as the distance from the main road leading to the CBD, distance from bus stations, distance from ambulances and schools, and distance from public parking The methodology involved evaluating model accuracy with metrics such as Mean Squared Error and Coefficient of Determination. Emphasizing LR's feature selection, significant predictors were identified, streamlining the model and offering insights into Prishtina's property price determinants, highlighting LR's effectiveness in feature selection in the context of emerging markets real estate valuation. The Lasso Regression model applied to Prishtina's apartment prices demonstrated moderate predictive accuracy, with an R² of 0.501, indicating it captures about half of the market's variability. Performance metrics, including MAE, RMSE, and MSE, underscore the tradeoff between model simplicity and precision. Despite lacking a direct feature importance ranking, Lasso's feature selection capability identifies key variables, offering insights into Prishtina's real estate market. This study highlights the potential of Lasso Regression in feature selection and regularization within emerging markets, showcasing its strength in simplifying complex datasets for enhanced interpretability and actionable insights, thus marking a significant contribution to the field by suggesting to integrate Lasso regression with other ML models for better predictive accuracy.