A machine learning approach to property valuation: a case study of Kampala Capital City Authority (KCCA)

Abstract

Introduction: The valuation of real estate properties is a critical task in the real estate industry, influencing decisions related to buying, selling, and investing. Traditional property valuation methods often rely on manual processes and expert judgment, which can be time-consuming and subjective. In recent years, machine learning techniques have shown promise in improving the accuracy and efficiency of property valuation by leveraging large datasets and advanced algorithms. By doing so, they can mitigate or eliminate the effects of decision-making bias and provide a more objective property valuation when trained on diverse datasets.

Purpose: The current property valuation system at KCCA is inadequate and this project aims to investigate the utilization of machine learning algorithms in property valuation, with the objective of developing a model that improves accuracy, efficiency, and decision-making in property valuation processes by providing precise property value estimates. Method: The study employs a dataset of 40,739 property records from Nakawa Division in Kampala, encompassing features such as property location, usage, building construction, size, amenities, rateable and gross values, among others that was analyzed using MS Excel and Python. The valuation model was developed and trained using Decision Trees, Gradient Boosting, Random Forest, and Linear Regression model machine learning algorithms applied to the dataset.

Results: This study demonstrates that the proposed model significantly expedites the process of property valuation compared to conventional methods. Furthermore, through model training, it was observed that tree-based models such as Decision Trees, Gradient Boosting, and Random Forest outperform the Linear Regression model, the previously chosen machine learning algorithm for this research. Among the tree-based models, the Decision Tree performed the best, followed by Gradient Boosting, and lastly, Random Forest, highlighting the overall superiority of tree-based models in accurately predicting rateable property values.

Conclusion: The findings of this study highlight the potential of machine learning in revolutionizing property valuation practices. By leveraging advanced algorithms and large datasets, machine learning can provide more accurate and efficient property valuations, benefiting both industry professionals and consumers.