An investigation of the performance of machine learning methods in the prediction of home sales to foreigners in turkey

Nowadays, forecasting techniques are gaining importance. Especially in recent years, the use of machine learning methods in forecasting has been increasing day by day. In this case, it isimportant to determine the most appropriate method by utilizing different methods. In recent years, the number of foreigners migrating to our country has increased considerably. Therefore, the desire to settle naturally increases the demand for housing sales. At this point, supply and demand in the housing sector must be kept in balance. For this reason, it is necessary to accurately forecast foreign demand for housing. In this thesis, it is aimed to reach the most appropriate prediction method by comparing the performances of four machine learning methods in the prediction of housing sales to foreigners. These methods are Multiple Linear Regression, Multilayer Neural Network, Decision Tree Regression, Random Forest Regression. In this study, the performances of the methods are compared using two different data sets. In the first dataset, there are 16 factors with a total of 127 months between 2013-2023, and in the second dataset, there are 20 factors with a total of 102 months between 2015-2023. In these two datasets, 4 different machine learning methods were applied. The performances of these methods are compared. As a result of the comparison of the methods, it was found that the Multiple Linear Regression method provided a better prediction than the other methods, with the Random Forest Regression model as the second method.