The real estate price prediction of US prediction based on multi-factorial linear regression models

Abstract

In every period, housing price prediction has always been a fascinating topic. Fluctuations in housing price are not only relevant to each individual resident but also to the politics and economy of the country. This essence of this research project is the usage of some real influencing factors to predict housing prices. In the Ames Housing dataset from Kaggle.com, five real factors that have a relatively strong correlation with housing prices are the overall material and finish quality, the above ground living area, the size of garage in car capacity, the garage area, and the total basement area. Based on these five real factors, two multiple linear regression models are constructed for predicting residential prices in Ames, Lowa, US. According to the analysis, when two independent variables are closely related, removing one of them does not necessarily reduce the fit of the model significantly, even if both independent variables are closely related to housing price. Therefore, choosing more appropriate variables is very important to increase the fit of the model. These results shed light on guiding further exploration of using more significant variables to find more accurate models to fit actual housing prices.