Issue: № 5, 2024
Doi: https://doi.org/10.37634/efp.2024.5.5
The paper develops an economic-mathematical model of housing prices in Dnipro city. Data for 2024 were collected, which included: Price per square meter (US dollar); Distance from the center (km); Total area (square meter); Number of rooms; Floor; Type of walls; Repair; Availability of autonomous heating; Height difference; View of windows; Guest bathroom; View of the ceiling; View of the hall floor; Guest area; Bedroom; Floor in the kitchen; Bathroom floor; Type of heating batteries; Ceiling height; m; Residential area (square meter); Kitchen area (square meter); Balcony/loggia; Air conditioners; Bathroom; Kitchen furniture; Bathroom furniture; Bedroom furniture; Furniture in the living room; Machinery; Availability of Wi-Fi; Bathroom area (square meter); Lighting devices in the kitchen; Lighting devices in the bathroom; Lighting devices in the bedroom; Lighting devices in the living room; Year of construction of the building. 24 classifiers were created to determine quantitative signs of qualitative characteristics of housing, in which class numbers start from 1. Based on these data, k-means clustering was carried out, as a result, 6 different classes were formed. An economic-mathematical model was created for each class, built on the basis of neural networks implemented in the Statistica software package. Each grid included three layers, and the hidden layer had at least 50-100 neurons. The quality of the model for each class ranged from 75% to 100%. Classification by the method of neural networks was preceded by a check of the level of correspondence of the previously found class of the object. It was recognized that in 70% of the objects, the class coincided with the significant class by the k-means method. The calculation of the price of a new apartment begins with the classification of its parameters with the determination of the number and the choice of the cluster to which it should be assigned. Next, the price is calculated according to the model of the corresponding cluster. Since for some clusters from 2 to 5 models are calculated, the average of the predicted price is found, and its standard serves as a possible border for trading between the buyer and the seller. Checking the accuracy of the model on data that was not included in the construction of the model showed an average deviation within 7 %.
Keywords : Housing prices in the Dnipro city, neural networks, cluster analysis, classification, regression model
References:
1. Molodchenko T.H., Tkachenko A.Yu. Real estate valuation. Methodical instructions for performing practical works in the discipline "Real Estate Appraisal". Kharkiv, 2011. 16 p. URL: eprints.kname.edu.ua/24845/1/Копия метод РГЗ оценка недвиж печ 2010.pdf (in Ukrainian).
2. Prysenko H.V., Ravikovych Y.I. Forecasting socio-economic processes. Kyiv, 2005. 378 p. (in Ukrainian).
3. How the DIM.RIA real estate evaluation service works: website. URL: https://mind.ua/publications/20179589-dom-ria-zapustiv-bezkoshtovnij-servis-ocinki-kvartir (in Ukrainian).
4. Uvecon PRO estimator calculator. URL: https://www.uvecon.ua/images/tinymce/file/Robotic%20calculator%20principle.pdf (in Ukrainian).
5. Buy an apartment in Dnipro. URL: https://dom.ria.com/uk/search?excludeSold=1&category=1&realty_type=2&operation=1&state_id=11&price_cur=1&wo_dupl=1&sort=inspected_sort&firstIteraction=false&city_ids=11&client=searchV2&limit=20&type=list&ch=242_239,247_252 (in Ukrainian).
6. K-means clustering: website. URL: https://en.wikipedia.org/wiki/K-means_clustering
7. Pistunov I.M. Neural nets and their application. Dnipro, 2024. 28 p. URL: https://ir.nmu.org.ua/bitstream/handle/123456789/163598/NeuroNets.pdf?sequence=1&isAllowed=y (in Ukrainian).