raimbekovm/bishkek-real-estate

🏠 Bishkek Real Estate Dataset

The largest open dataset of apartment listings from Kyrgyzstan

Dataset Files • Quick Start • Features • Benchmarks • Use Cases

---

📊 Overview

Comprehensive dataset of 8,821 apartment listings from Bishkek, Kyrgyzstan with:

• 💰 Prices in USD and per m²
• 📍 GPS coordinates for 99.6% of listings
• 🏗️ Building data: year built, house type, floors
• 🏢 Residential complex info (name, class, developer)
• 📝 Descriptions in Russian

Why This Dataset?

Feature	Value
🌍 Unique geography	First ML-ready real estate dataset from Central Asia
📈 Rich features	60 columns including GPS, building specs, complex data
🎯 ML-ready	Clean, preprocessed, with benchmark models
🆓 Open license	CC-BY-4.0, free for commercial use

---

📁 Dataset Files

├── listings.csv                    # Main dataset (8,821 rows, 60 columns)
├── listings_with_rc.csv            # With residential complex data
└── notebooks/
    ├── bishkek_v3_best.ipynb       # Best model (MAE $121.71/m²)
    └── bishkek_cv_model.ipynb      # CV experiment with ResNet-50

---

🚀 Quick Start

from datasets import load_dataset

# Load dataset
dataset = load_dataset("raimbekovm/bishkek-real-estate")
df = dataset["train"].to_pandas()

# Explore
print(f"🏠 Apartments: {len(df):,}")
print(f"💰 Median price: ${df['price_usd'].median():,.0f}")
print(f"📐 Median $/m²: ${df['price_per_m2'].median():,.0f}")

With Pandas

import pandas as pd

df = pd.read_csv("hf://datasets/raimbekovm/bishkek-real-estate/listings.csv")

---

📋 Features

Price & Size

Column	Type	Description
price_usd	float	Price in USD
price_per_m2	float	Target variable — Price per m² in USD
rooms	int	Number of rooms (1-6+)
area	float	Total area in m²

Location

Column	Type	Description
latitude	float	GPS latitude (WGS84)
longitude	float	GPS longitude (WGS84)
district	str	Neighborhood (287 unique)

Building

Column	Type	Description
floor	int	Apartment floor
total_floors	int	Building total floors
year_built	int	Construction year
house_type	str	монолит / кирпич / панель

Residential Complex

Column	Type	Description
jk_name	str	Complex name
jk_class	str	комфорт / бизнес / элит
developer	str	Developer company

---

🏆 Benchmarks

Model Performance

Model	MAE ($/m²)	R²	MAPE
Ensemble + Optuna	$121.71	0.76	8.2%
XGBoost	$128.45	0.73	8.9%
CatBoost	$126.12	0.74	8.5%
LightGBM	$130.23	0.72	9.1%

Key Finding

📌 Image embeddings (ResNet-50) do NOT improve price prediction. Tabular features alone achieve best results.

---

💡 Use Cases

1. Price Prediction — Build regression models to estimate apartment prices
2. Market Analysis — Analyze price trends by district, building type
3. Geospatial ML — Use GPS coordinates for location-based features
4. NLP — Extract features from Russian descriptions
5. Anomaly Detection — Find underpriced/overpriced listings
6. Academic Research — Study emerging real estate markets

---

📈 Dataset Statistics

Metric	Value
Total records	8,821
Features	60
With GPS coordinates	99.6%
With year_built	73.1%
Linked to residential complex	48.9%
Price range	$23K — $1.5M
Price per m² range	$400 — $3,500

---

📍 Data Source

Data collected from house.kg — the largest real estate platform in Kyrgyzstan.

Collection period: December 2025 — January 2026

---

📜 License

This dataset is released under CC-BY-4.0.

You are free to:

• ✅ Use for commercial purposes
• ✅ Modify and adapt
• ✅ Distribute

With attribution to the author.

---

📚 Citation

@dataset{bishkek_real_estate_2026,
  title={Bishkek Real Estate Dataset},
  author={Raimbekov, Murat},
  year={2026},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/raimbekovm/bishkek-real-estate}
}

---

👤 Author

Murat Raimbekov

• GitHub: @raimbekovm
• LinkedIn: Murat Raimbekov
• Email: murat.raimbekov2004@gmail.com

---

⭐ If you find this dataset useful, please give it a like!