CEU Electronic Theses and Dissertations, 2022
| Author | Ayobi, Ghazal |
|---|---|
| Title | Transaction Price on Property Market |
| Summary | The aim of this project is to evaluate transaction price of properties to build a prediction model. The main goal of the project is to predict prices of flats in capital region and other major cities. The key variable is transaction price. For data analysis and feature engineering, first I created several groups of the predictors based on the variable importance. Afterwards, I applied various linear regressions on the grouped variables, creating five OLS models with different difficulty levels. As a result, the best performing regression is chosen for the further analysis in each group of data sets. Based on the 5-fold cross-validation RMSE best models were chosen for the capital region and cities. These models have the lowest RMSE in the test sets. After selecting the best OLS model, I evaluated the data using machine learning models. A machine learning algorithm is a methodology through which an AI system undertakes a task and predict the value based on the given dataset. It is important to run and evaluate different models and methodologies for a given data set, as this assist to understand the dataset in broader perspective. In order to predict flats transaction prices, I used CART, Random Forest and GBM algorithms. Based on the work and holdout data performance and RMSE, ordinary least square method provided the best performance across different algorithms. |
| Supervisor | Gabor Bekes |
| Department | Economics MSc |
| Full text | https://www.etd.ceu.edu/2022/ayobi_ghazal.pdf |
Visit the CEU Library.
© 2007-2021, Central European University