House price forecasting based on ensemble learning

Abstract

During the last decades, the housing market has played a significant part in the development of contemporary civilization. They enable the deployment of economic resources. Changes in housing prices reflect changes in the market. With extensive data processing capabilities in various domains, machine learning is also widely employed in the financial industry such as housing price prediction, optimum investment, financial information processing, and execution. financial trading tactics. Therefore, house price prediction is viewed as one of the most popular and lucrative topics in the financial business. In this research, I propose Ensemble Learning in the Machine Learning method in house price prediction. Ensemble learning algorithms based on boosting (Light Gradient Boosting, Gradient Boosting Regressor, and Extreme Gradient Boosting), Random Forest is an example of a bagging method are used and compared with stacking. Besides, I also use some regression models such as Ridge, and Lasso to predict and compare. I tested residential homes in Ames, Iowa with 79 properties per home. Experimental results show that our proposed Ensemble Learning method can achieve good results in predicting house prices compared to many conventional prediction models. The thesis also proposes to build a web application to visualize research results and support users to predict house prices from the properties of the house, which strongly affect the value of a house. Experimental results show that the proposed model achieves good results on the datasets used for training and testing on all measures: root mean square error (RMSE), K-fold cross-validation.