Author(s): Ruth Wangari Muchai

Abstract: The real estate sector in Kenya plays a pivotal role in economic development yet traditional valuation methods often struggle to accurately reflect dynamic market conditions. This research explores the transformative potential of advanced data-driven techniques to enhance real estate valuation accuracy focusing on residentials within Nairobi metropolis for the period between 2009 to 2024. The research assesses the effectiveness of supervised machine learning models. The models evaluated included linear regression, lasso regression, ridge regression, gradient boosting, and random forest algorithms. Data was sourced from Highlands Valuers firm covering 521 records of residential properties. The research employed exploratory data analysis and model development. Lasso regression emerged as the best model due to its significant improvements in RMSE and R-squared values after hyperparameter tuning and regularization. This demonstrated superior accuracy and predictive capability. The most significant features were comparable sales, socio-economic status of the area, built-up area, number of floors, total number of rooms, rent, building age, residential development type, number of bedrooms, neighborhood characteristics, environmental considerations, and proximity to amenities. The research underscored the potential of machine learning to revolutionize real estate valuation practices in Kenya. The research outcomes are expected to inform stakeholders, policymakers, and practitioners and foster more informed decision-making, improve sector stability, and support sustainable growth in the Kenyan real estate market.

DOI: 10.22271/maths.2024.v9.i4a.1769

Pages: 49-60 | Views: 191 | Downloads: 36

Download Full Article: Click Here