earthquake.kz — Machine Learning Seismic Forecasting System for Central Asia
🌍
earthquake.kz
Seismic Forecasting System
LSTM
CatBoost
Central Asia
Analytics
About
Forecast Risk
Low (< 0.5)
Medium (0.5 – 2)
High (2 – 5)
Very High (5+)
Click on map or cell to predict
LAT
—
LON
—
Болжам
Дерек
Тарих
About earthquake.kz
✕
ML-based seismic forecasting system for Central Asia · LSTM + CatBoost + DNN
System
Framework
Laravel 11 + PyTorch
Data source
USGS 1973 – present, M≥3.0
Region
Central Asia (35–55°N, 46–87°E)
Target
Next-day event count (Poisson)
Train split
≤2022 train / 2022–2023 val / 2023+ test
LSTM Architecture
Input
7 days × 11 features
↓
LSTM
hidden_size = 32
↓
Dropout
p = 0.2
↓
Linear
32 → 1 (log λ)
↓
Output
exp(log λ) = λ events
Features (11)
Rolling 7d
count, mean_mag, max_mag
Rolling 30d
count, mean_mag, max_mag
Rolling 180d
count, mean_mag, max_mag
Static
dist_to_fault, mean_depth
Dataset Analytics
✕
USGS seismic catalog · Central Asia · 1973–2026
—
Total Events
—
Period
—
Avg Magnitude
—
Max Magnitude
Events per Year
Magnitude Distribution
Monthly Seasonality
Depth Distribution
Model Comparison
Loading…