| Keywords [eng] |
Cross-sectional data, OLS Regression, Support Vector Regression, K-Neighbors Re- gression, Decision Trees Regression, Gradient Boosting Regression, MLP Regression, Car price |
| Abstract [eng] |
This master thesis evaluates the ability of various regression methods to predict car market price. It describes and assesses different models such as OLS Regression, Support Vector Regression, K-Neighbors Regression, Decision Trees Regression, Gradient Boosting Regression, and MLP Regression. The study outlines the steps for data preparation, modeling and hyperparameter tuning, while also proposing a modeling framework for predicting prices for each car brand separately. Moreover, distinct from similar studies, this research includes monthly dummy variables along with the more conventional car features. The study finds that Decision Trees Regression emerges as the most effective method, demonstrating high performance while also being time-efficient. In addition, Horse Power and Age at Sale stand out as the features having the most importance. However, the results could be further improved by including more independent variables such as trim level or accident history and using higher quality data with less missing information. Moreover, integrating extended time series data with macroeconomic indicators could also provide a clearer picture of the variations in car prices. |
