Institutional Repository of Vilnius University
Nekilnojamo turto kainų rinkimas, analizė ir prognozavimas /
Abstract
dc:descriptionReal estate prices are a very important topic these days. Existing real estate portals provide only a limited amount of historical price changes and future trends. Also these portals main objective is to sell, which means it is not relevant for them to display such data. This paper describes a tool that was developed for automated data collection, analysis and forecasting of real estate prices. A lot of attention is spent on describing what are time series forecasts models and neural networks forecasting models. Also, this paper examines ARIMA, SARIMA and LSTM model working principles more deeply. The data for forecasting was collected during the time period from 2023-10 to 2024-05. Average flat in Kaunas price per square meter is used for forecasts. After preparing the data for further use, analysis was made of how each parameter affects each model prediction accuracy. The accuracy of models was also compared. The Long short-term memory model shows more accurate results (1.17% error) compared to the time series models ARIMA (2.43% error) and SARIMA (4.15% error). Also, this model is not so sensitive to data fluctuations and non-stationary data. ARIMA and SARIMA can only be used to identify a general trend, while LSTM can be used for more accurate forecasts that pay attention to details.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Aleknavičius, Tautvydas,
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD210646154&prefLang=en_US