Predicting Rental Value Of Residential Properties In Colombo City Using Machine Learning Techniques

Abstract

"Real estate is a well stabilized and an actively engaging sector throughout the globe with standardized set of guidelines that are initiated towards the benefit of both landlords and the tenants. This is not the case to be observed in Sri Lanka and there is a doubt whether there is a proper framework to determine and valuate a price for properties. Sri Lanka being a third world developing country, people find it hard to construct their own house or property which makes them vulnerable towards renting a property. The most common aspect is renting out residential properties and majorly renting out in the capital city due to many reasons. With the increase in demands for jobs within the Colombo city regions, people tend to find residential properties for rent within this area which is more convenient for them in terms of accessibility and the factor of heavy traffic. Also, with the urbanization and the development of resources in the Colombo city limits, temporary migration is a common factor which has been influencing the determination of rental values of properties. Hence, in this study major cities within the Colombo city limit were selected to ensure that there is no major deviations are observed in terms of the rental value. Both statistical and Machine Learning approaches are used in this study to construct models. Machine Learning techniques like Linear Regression, Ridge Regression, Lasso Regression, Partial Least Square Regression, K-Nearest Neighbor Regression, Decision Tree Regression, Random Forest Regression and Gradient Boost Regression was used. Each of these models were trained and tested and the evaluation matrices were recorded. Evaluation of the models were done in terms of R2, RMSE and Cross Validation CV. Hyper parameter tuning was used to determine the best combination of parameters in order to construct the model."