FORECASTING THE VOLUME OF RESIDENTIAL REAL ESTATE SALES IN A NEURAL NETWORK BASIS

Abstract

While developing the economic part of a construction project, a cash flow model is built to consider all the key factors that affect the overall project management system. An important component in building a cash flow model is knowing the volume of future sales. Forecasting the volume of sales allows you to predict the income from the implementation of the project and assess its profitability. Currently, construction  companies assess volume of sales expertly, and the results of the forecast depend on the expert’s experience. In order to improve the efficiency of building a cash flow model, the paper pro-poses a neural network (NN) model for forecasting of the volume of real estate sales considering market factors.
The model is built on the basis of the Loginom analytics platform, trained and has good predictive properties. The average relative forecast error is 5.21%. The model considers statistically significant external and internal factors affecting the volume of real estate sales under shared-equity construction in the Chelyabinsk region market.