Exploratory data analysis of apartment listings (Yandex Real Estate archive): what drives housing prices in St. Petersburg and its suburbs. Bilingual project — full analysis available in English and Russian.
Key findings: apartment price is driven primarily by total area and location — median price 4.65M ₽; half of all apartments sell within 95 days; publication day and month have no meaningful effect on price.
The dataset is an archive of apartment listings for St. Petersburg and nearby settlements. The goal is to identify the parameters that determine market value — groundwork for an automated system that flags anomalies and fraudulent listings.
23,699 listings (23,613 after cleaning) listings with two types of features: user-submitted (price, area, rooms, floor, ceiling height) and map-derived (distances to city center, airport, parks and ponds).
Preprocessing: missing values filled by feature type (medians for numeric, explicit categories for strings), implicit duplicates in settlement names unified, data types converted, outliers filtered by quantiles and IQR.
Engineered features: price per m², publication day of week / month / year, floor type (first / last / other), distance to city center in km.
Price factors. Total area is the strongest price driver; living area, kitchen area and number of rooms follow (living area correlates with total area at 0.88, so their effects overlap). Price per m² rises with total area (correlation 0.66) — larger apartments are more expensive per meter as well. Balconies, publication day and month show no meaningful effect.
<
Location. Prices peak in the city center (0–5 km) and stabilize at 5–25 km. Suburban resort areas (Lisy Nos, Zelenogorsk) show elevated price per m² despite the distance — nature premium at work.
Market dynamics. Median time on market — 95 days (mean 168 days, skewed by long-tail listings). Sales within 30 days can be considered fast; listings over 100 days warrant a closer look — potential overpricing or property issues.
Python · pandas · NumPy · Matplotlib · Seaborn · Jupyter
git clone https://github.com/foxypandas/real-estate-analysis.git
cd real-estate-analysis
pip install -r requirements.txt
jupyter notebook notebooks/real_estate_eda_en.ipynbThe dataset is not included in the repository.
├── notebooks/ # analysis notebooks (EN and RU versions)
├── images/ # charts and visualizations
└── requirements.txt