Predicting the Price of Cars in Türkiye Using Machine Learning Techniques


Thesis Type: Postgraduate

Institution Of The Thesis: Marmara University, Institute of Social Sciences, Department of Business Administration (Eng), Turkey

Approval Date: 2024

Thesis Language: English

Student: YASİN ONUR ABAT

Supervisor: Hüseyin Ekizler

Abstract:

The aim of this research is to estimate the prospective six months sales price of vehicles in the Automobile and SUV categories in Turkiye. This research was done with the information obtained from the classifieds posted on sahibinden.com. By using the vehicles’ brand, model, package type, model year, body type, gear type, fuel type and mileage characteristics for the future six months sales prices of the vehicle was estimated over the past selling prices of vehicles. For this regard, six different algorithms (ANN, ARIMA, SARIMA, XGBOOST, TBATS, HOLT WINTERS Forecasting) were studied in this research.  According to the success of these algorithms, the best algorithm which is HOLT WINTERS was selected. The results of HOLT WINTERS were announced as output of the research. As a result of this research, basically the  answer to the following question was sought; What will be the price of the vehicles to be sold/desired to be purchased within the time frame six months after the current date? By answering this question, it was created an environment to determine the value of the vehicles more accurately and reliably during vehicle purchase/sale transactions. Thus, sellers and buyers will be able to evaluate their situation economically better and be able to perform their transactions better.