Models for forming the market value of a real estate

Abstract

The article deals with the problem ofa high-quality and quick assessment of real estate. It is proposed to use machine learning models and technologies to create a real estate evaluation service based on the entered characteristics. Two groups of parameters are used in the proposed models: characteristics of the area in which the real estate is located, and characteristics of the real estate object. Since several objects are located in the same region, data about the region will often be repeated. Two models are implemented: a model with two neural networks and a model with a neural network and a random forest. The MSE is chosen as the loss function. Data on already concluded agreements were used to evaluate the effectiveness of the proposed models. The developed service is aimed at owners of residential real estate who want to sell or rent it, and buyers who want to check the fairness of the price. To form a value, users are required to provide a detailed description for an accurate result, but some of the parameters are optional to obtain an approximate value for guidance. In addition, unlike other implementations, this service offers to choose a period during which the seller wants to find a buyer. Ref. 7, pic. 3, tabl. 1.

Keywords

• machine learning
• cost forecasting
• cloud computing