House Price Prediction: A Case Study for Istanbul

Abstract

AbstractOne of the biggest investments people make is undoubtedly the purchase and sale of real estate. The most important of the real estates is the purchase and sale of housing. The housing market, which is also an indicator of social wealth, is one of the important elements of the economy. Changes in house prices affect not only the housing market but also the economy indirectly. For this reason, the correct determination and estimation of the financial values of the houses are of great importance in terms of the stability, reliability, and sustainability of the housing market. Hedonic price model (HPM) and artificial neural networks (ANNs) are the most widely used methods in estimating housing prices, which have a very heterogeneous structure. HPM is an estimation method based on linear regression analysis that explains the relationship between dependent and independent variables with linear relationships and requires some assumptions, while ANN is a method without limitations. In this study, a random sample was selected by creating a list of houses sold in Istanbul between 2015 and 2019. Using this dataset, house sales prices were estimated with HPM and ANN models. For this purpose, different criteria have been taken into account, especially in the region where the residences are located. From the results obtained, it was observed that ANN gave more consistent results than HPM.KeywordsArtificial neural networkHedonic price modelMultilayer perception