The thesis introduces several methods of real estate price modelling suitable either for prediction of the housing prices or for exploring the relationships between the price and its determinants. We compared the conventional linear regression approach to the tree-based methods of machine learning. The comparison analysis on the dataset of 28 019 apartments in Prague suggests that regression trees (especially the Random forest) yield a higher accuracy in the price prediction. Another objective was to examine the effects of location attributes (especially its accessibility and environmental quality) on the prices of nearby apartments. To address the spatial interactions in the geographical data, we employed three spatially conscious models to achieve more reliable results. The local analysis performed with the geographically weighted regression confirmed the presence of spatial heterogeneity and described the price effects relative to the location. In some areas, an increase of 100 meters in distance from the nearest metro station and the nearest park are associated with a decrease in the apartment prices by 644 CZK/m2 and 916 CZK/m2 , respectively. These findings are especially important for the apartments near the stations of the new metro line, which is currently in construction.
| Identifer | oai:union.ndltd.org:nusl.cz/oai:invenio.nusl.cz:435251 |
| Date | January 2020 |
| Creators | Charvát, Ondřej |
| Contributors | Polák, Petr, Hejlová, Hana |
| Source Sets | Czech ETDs |
| Language | English |
| Detected Language | English |
| Type | info:eu-repo/semantics/masterThesis |
| Rights | info:eu-repo/semantics/restrictedAccess |
Page generated in 0.0022 seconds