Predicting Residential Property Valuation in Major Towns and Cities on Mainland Fiji

Abstract

The real estate sector plays a pivotal role towards the economy of every nation. Transaction of properties heavily rely on the valuation price as determined by the appraisers who use many variations of techniques to determine the valuation of respective properties. This value is used by investors, sellers, intermediary agencies such as real estate agencies and financial institutions as well as government entities. However, the value determined by the appraisers is just an approximation, excluding the accuracy and error rate with respect to the actual price. This study aims to integrate the capabilities of machine learning models and algorithms towards determining valuation price. Four algorithms were selected, namely, Multiple Linear Regression, Decision Tree, Random Forest, and Artificial Neural Network, for supervised learning against training and testing of a dataset acquired from real estate agencies based on residential properties in major towns and cities on mainland Fiji. Results show that Random Forest and Artificial Neural network produce high levels of accuracy based on Mean Absolute Error (MAE), Mean Squared Error (MSE) and Root Mean Squared Error (RSME) values. The study has significantly contributed towards developing insights to developing accurate models which could enable users access to valuation of properties based on the input of property features through machine learning predictions.