Building a Predictive Model for Predicting Real Estate Prices Based on the Generated Database

Abstract

The work is devoted to solving the current problem of forecasting real estate prices by building a predictive model based on the generated database of real estate in Moscow, posted on the Move Real Estate website. Existing machine learning methods for solving the forecasting problem are considered and one of them is applied - multiple linear regression. A regression analysis of the obtained results of solving the forecasting problem was carried out. Eleven independent variables are considered as control parameters. The influence of the variables taken into account when constructing the model on the results of solving the problem of forecasting real estate prices was studied. It was determined which of the independent variables have the greatest impact on the results of the model. To improve the quality of the model, preprocessing and standardization of features were carried out. Identification of outliers and omissions of values was carried out during the formation of the database. The coefficients of the multiple linear regression model were determined using the least squares method. To assess the quality of the model, the following model parameters are analyzed: R-squared, adjusted R-squared, p-value. The result of constructing a predictive model is the resulting regression equation. The application of the resulting equation can be used to subsequently take into account specific characteristics when solving the problem of forecasting real estate prices. The work shows the advantages of using this method and the prospects for applying the obtained result.

Keywords

• machine learning
• regression analysis
• multiple linear regression
• least squares method
• regression equation
• r-squared
• adjusted r-squared
• p-value