House price prediction using machine learning

Abstract

The role of the real estate industry in economic development and social progress reflects the economic well-being of individuals and regions. With the increase of people's income level, the demand for housing is also increasing. Therefore, making a more accurate house price forecast will help people make the most correct strategy to buy a house when they need it. This study focuses on house price prediction in King County, Washington, a diverse real estate market. Leveraging machine learning models such as linear regression, random forest, neural networks and XGBoost, these supervised learning models are used to delve into house price forecasting. This research includes random forest and XGBoost, are implemented using Scikit-Learn tools. Besides, the Feedforward Neural Network is introduced with the drop out layer in order to reduce the occurrence of model fitting situations. The findings reveal that XGBoost achieves the highest accuracy, making it well-suited for precise price predictions. Additionally, the research identifies grade, sqft_living, and latitude as the three most influential features significantly affecting house prices within the dataset.