ExoHunter: Exoplanet Detection and Classification System with Machine Learning 🔭

Overview

ExoHunter is a Data Science project that leverages machine learning to detect and classify exoplanets from telescope data, focusing on light curves from NASA's Kepler and TESS missions.

By analyzing stellar flux to identify planetary transits, this project combines astrophysical passion with advanced data engineering and ML techniques, making it a standout addition to any Data Science portfolio.

---

✨ Why ExoHunter is Unique

• Astrophysical Storytelling: Inspired by the hunt for "cosmic mysteries," ExoHunter not only classifies exoplanets (Confirmed, Candidate, False) but also detects rare anomalies in light curves, playfully exploring "exotic" signals like those from Tabby's Star.
• End-to-End Pipeline: From raw data ingestion to interactive dashboards, it showcases full-cycle Data Science: ETL pipelines, feature engineering, ML modeling, and deployment.
• Real-World Relevance: Demonstrates skills in time-series analysis, imbalance handling, and Explainable AI, directly applicable to industries like space tech, fintech, and IoT analytics.

---

🎯 Project Goals

1. Data Processing: Build a scalable ETL pipeline to clean and transform raw telescope data.
2. ML Classification: Train models to classify exoplanet signals with high accuracy (target: $AUC > 0.95$, $recall > 0.9$ for confirmed exoplanets).
3. Anomaly Detection: Use unsupervised learning and Bayesian inference to flag unusual signals.
4. Deployment: Deliver an interactive web app for users to upload light curves and view predictions.

---

🛠️ Tech Stack

Category	Key Tools
Data Engineering	PySpark/Hadoop (big data), Pandas (initial ETL), PostgreSQL (structured metadata), MongoDB (raw light curves).
Machine Learning	Scikit-learn (Random Forest, XGBoost), TensorFlow (CNNs for light curves), PyMC (Bayesian inference).
Visualization	Matplotlib/Seaborn (exploratory plots), Streamlit/Superset (dashboards).
Deployment	FastAPI (API), Dokploy on AWS Lightsail (hosting).
Other Tools	Jupyter (EDA), Git (version control).

---

📂 Datasets

• Primary Source: NASA's Kepler Exoplanet Search Results (Kaggle) with $\sim9,500$ stellar observations, including labels (Confirmed, Candidate, False) and features like transit depth and period.
• Future Sources: TESS light curves from MAST, NASA Exoplanet Archive API for real-time data.

---

🌳 Project Structure

---

ExoHunter/
├── data/               # Raw and processed datasets
├── notebooks/          # Jupyter notebooks for EDA
├── scripts/            # Python scripts for ETL and ML
├── models/             # Trained ML models
├── visualizations/     # Plots and dashboards
├── requirements.txt    # Dependencies
└── README.md           # Project documentation

🚀 Current Progress and Next Steps

Current Progress

• Day 1: Set up GitHub repo, installed dependencies, downloaded Kepler dataset, and performed initial EDA with basic visualizations (e.g., period distributions).

Next Steps

1. Data cleaning and feature engineering (e.g., $SNR$, transit duration).
2. Training a baseline Random Forest model.

Future Roadmap

• Automate data ingestion via NASA API.
• Implement CNNs for light curve analysis.
• Deploy a web app with real-time predictions.
• Generate narrative PDF reports (e.g., "Discovery Report: A Hidden World Detected").

---

⭐ Why It Stands Out for Employers

ExoHunter bridges astrophysical curiosity with enterprise-grade Data Science:

• Scalable Pipelines: Handles large-scale time-series data, relevant for IoT or financial analytics.
• Advanced ML: Tackles class imbalance and complex patterns, showcasing skills in model robustness.
• Creative Narrative: The "cosmic mystery" angle makes it memorable, differentiating it from generic retail or finance projects. --

💻 Installation

To explore ExoHunter locally, follow these steps:

1. Clone the repo:

   git clone https://github.com/betoalien/ExoHunter.git

2. Create a virtual environment:

   python -m venv exohunter_env
   # Activate environment (Linux/macOS):
   source exohunter_env/bin/activate
   # Activate environment (Windows):
   # exohunter_env\Scripts\activate

3. Install dependencies:

   pip install -r requirements.txt

4. Run EDA notebook:

   jupyter notebook notebooks/eda.ipynb

---

🤝 Get Involved

Interested in ExoHunter or have questions about the project? Feel free to reach out to me!

Email: conect@albertocardenas.com

Contributions, feedback, or inquiries about Data Science opportunities are welcome!

---

Built with passion for astrophysics and data-driven discovery by Alberto Cardenas.