This project develops and deploys a robust machine learning model to nowcast the likelihood of significant earthquakes (Magnitude 6.0+) in California within a 30-day horizon. Leveraging historical seismic data and advanced feature engineering, the system provides automated, data-driven insights into current seismic risk.

• Comprehensive Data Acquisition: Fetches M2+ earthquake data from the USGS FDSN client for California (1990-Present) with robust chunking to handle API limits.
• Advanced Feature Engineering: Calculates both regional and fine-grained spatial seismic features (seismicity rate, b-value, mean magnitude, etc.) from sliding time windows.
• XGBoost Classification: Utilizes a highly optimized XGBoost model for predicting large earthquake probabilities.
• Class Imbalance Handling: Employs SMOTE to balance the highly imbalanced earthquake dataset during training.
• Optimal Thresholding: Identifies and applies an optimal probability threshold for binary predictions, balancing precision and recall.
• Automated Prediction Pipeline: A dedicated script for real-time (or near-real-time) predictions, loading the trained model and processing the latest seismic data.
• Automated Scheduling: Configured for automated execution via system schedulers (cron/Task Scheduler).
• Robust Alerting: Includes logging to file and email notifications for predicted large earthquake events.

• Python 3.x
• ObsPy: For seismic data acquisition.
• Pandas & NumPy: For data manipulation and numerical operations.
• Scikit-learn: For machine learning utilities (train/test split, GridSearchCV, metrics).
• Imbalanced-learn (imblearn): For handling class imbalance (SMOTE).
• XGBoost: The core machine learning model.
• joblib: For model serialization.
• Matplotlib & Seaborn: For data visualization.
• smtplib & email.mime.text: For email notifications.

• Python 3.x installed.
• Git installed.
• A GitHub account (for pushing your code).
• VS Code (recommended IDE).

Clone the repository:

git clone https://github.com/oye-bobs/siesmicanalyzer
cd siesmicanalyzer

Create and activate a virtual environment (recommended):

python -m venv .venv
# On Windows:
.\.venv\Scripts\activate
# On macOS/Linux:
source ./.venv/bin/activate

Install dependencies:

pip install obspy pandas numpy scikit-learn imbalanced-learn xgboost matplotlib seaborn

Before running, you might need to adjust some configurations:

• CATALOG_FILE: Name of the CSV file to store the earthquake catalog.
• FORCE_FETCH: Set to True for the first run to download the full catalog.
• start_time_full_catalog: Defines the start date for data fetching (currently "1990-01-01").
• minlatitude_region, maxlatitude_region, minlongitude_region, maxlongitude_region: Define the geographical area of interest.
• target_magnitude: The magnitude threshold for "large earthquakes" (currently 6.0).
• prediction_horizon_days: The future window to predict large earthquakes (currently 30 days).

• OPTIMAL_THRESHOLD: Must be set to the exact optimal F1-score threshold found during training.
• Email Configuration: Update EMAIL_SENDER, EMAIL_PASSWORD, and EMAIL_RECEIVER for alerts.

python nowcasting_pipeline9.py

python predict_earthquake.py

crontab -e

Add a line:

0 3 * * * /path/to/your/venv/bin/python /path/to/your/project/predict_earthquake.py >> /path/to/your/project/prediction_log.txt 2>&1

• Search for "Task Scheduler"
• Create a "Basic Task"
• Set the schedule and point to your Python interpreter and script.

• True Positives: 18
• False Positives: 0
• False Negatives: 4
• Precision (Large Quake): 1.00
• Recall (Large Quake): 0.82
• F1-score (Large Quake): 0.90
• ROC-AUC Score: 0.9982
• Precision-Recall AUC: 0.94

• Advanced Feature Engineering (e.g., stress accumulation metrics).
• Time Series Cross-Validation (e.g., walk-forward validation).
• Ensemble Methods (LightGBM, CatBoost).
• Web Dashboard for visualization and interaction.

Feel free to fork this repository, open issues, or submit pull requests.

MIT License.

Your Name - [email protected]