House price prediction modeling using machine learning techniques: a comparative study
From Firenze University Press Journal: Aestimum
Ayten Yağmur, Department of Labour Economics and Industrial Relations, Akdeniz University
Mehmet Kayakuş, Department of Management Information Systems, Akdeniz University
Mustafa Terzioğlu, Accounting and Tax Department, Akdeniz University
Human needs are endless, however, some basic needs such as nutrition, shelter and protection should be first met for the continuation of their lives. Housing need is a multidimensional problem that is necessary for people’s shelter, health, security and various socio-cultural needs. People want to buy a house in order to have their own house only when their welfare reaches a certain level. At this stage, the important thing is to choose a house that will meet the budget they have and the needs of their family members. In this respect, affordable housing prices are very important for households. Housing suppliers prioritize the needs of these households while designing the house they will produce architecturally. It is very important that the house to be produced meets these needs appropriately and that it is built in the right location in terms of the costs to be incurred. It is of vital importance for these institutions to deter-mine the housing price correctly, to meet these costs, to sell the produced houses easily and to achieve a desired amount of profit margin. Because the institutions that supply housing make huge capital investments and the wrong construction projects that cannot be sold cause these institutions to go bankrupt very quickly. Banks, mortgage and real estate companies that provide hous-ing financing allocate loans to households that demand housing based on the housing price and the appraisal valuation they will make.
Therefore, the creation of an effective and effective credit policy by these financial institutions directly depends on the accurate price pre-diction of the house. Since the maturities of these loans will be medium and long term, incorrect loan allocation will reduce their assets and reduce their direct return on assets (ROA), because of these companies making an inefficient use of assets on their balance sheets in the long run. Thus, the main deciding factor for the three important actors in the housing market is the sale price of the house. The factors that determine the sale price of the house are primarily the basic features of the house. The first of them is the location of the house. In general, it is seen that houses are built according to the lower, middle- and upper-income groups and their needs depending on the features of the location (by the sea, in the forest, distance from the city center, school, hospital, religious places, and proximity to organized industrial zones, which are production zones, etc.). Another factor is the volume and situation of the house. The usable size and the number of rooms of the house are a direct factor for the selling price due to both the demographic char-acteristics of the demanding households and the cost of the housing to be built. Furthermore, the fact that the house sold is a new or secondary house directly affects the firm sale price of the house in high-type houses. Moreover, whether the house is designed as a complex of buildings (security, pool, Spa, gym, etc.) is a determining factor on the sales price of the house. These variables, which we describe as micro-var-iables, were made into a model in the study. With this model, it was attempted to predict the house prices using machine learning methods, which are among the advanced prediction techniques. It is considered that the obtained results will contribute to correct pricing in terms of housing suppliers, mediators in house sales and institutions that provide financing. The results of the model created in the study are also important in terms of an effective and active housing market. Especially in housing markets where price fluctuations are high and there are housing supply and demand imbalances, the use of advance price prediction mechanisms will ensure the proper operating of the markets.
The use of three different machine techniques in the study and especially the testing of the support vec-tor regression technique in this regard differs from simi-lar studies in the literature. The aim of this study is to create a model that can accurately predict the housing prices in the locations in the portfolios of the institu-tions that offer housing and mediate its sale. Testing the success of the designed model using machine learning methods is the second main objective of the study. At the same time, it is aimed to be an exemplary reference study for more appropriate housing production planning by considering the preferences of those who supply housing and those who demand it. To achieve these goals, the main hypothesis of the study is that the variables that reveal the characteristics of the house in the estimation of housing prices will be successfully predict-ed using machine learning methods.In the second section of the study, reference was made to the studies on the basic dynamics affecting housing prices. In addition, studies using machine learn-ing and other methods for housing prices are includ-ed. Section 3 describes the model of the study and the machine learning methods used by focusing on the data set of this model. In section 4, the results obtained by machine learning are included in the study and these findings are discussed. Section 5 presents the conclu-sions drawn from the study and the policies and recommendations drawn from these conclusions.
DOI: https://doi.org/10.36253/aestim-13703
Read Full Text: https://oaj.fupress.net/index.php/ceset/article/view/13703
