🪐 Exoplanet Habitability Analysis & KOI Classification

An end-to-end data science project built using Kepler exoplanet data. This project includes data preprocessing, feature engineering, habitability scoring, exploratory analysis, classification modeling, and a fully interactive Streamlit dashboard.

---

📌 Project Overview

This project aims to explore and classify Kepler Objects of Interest (KOIs) to determine their potential habitability and disposition using scientific metrics and machine learning.

The complete pipeline includes:

1. 🧹 Data Cleaning & Preprocessing
2. 🧪 Exploratory Data Analysis (EDA)
3. 🛠️ Feature Engineering
4. 🌍 Habitability Scoring System
5. 🎯 KOI Disposition Classification (Machine Learning)
6. 📊 Visualizations

---

🌐 Project Workflow

1. 🧹 Data Cleaning

• Loaded KeplerExoRaw.csv
• Removed columns with excessive missing values
• Handled missing data via imputation
• Converted types and removed duplicates

2. 🛠️ Feature Engineering

• Created features like:
  • pos_diff_mdec, pos_diff_msky: positional differences
  • avg_err_mdec: average error
  • total_pos_diff: total spatial noise
  • Encoded categorical flags
  • Extracted vetting year

3. 📈 Exploratory Data Analysis

• Distributions of planetary and stellar features
• Correlation heatmaps
• Trends across vetting years
• Comparison of features across KOI classes

4. 🌍 Habitability Scoring System

Planets scored from 0 to 3 based on:

Feature	Ideal Range
Radius	0.5 – 1.5 Earth radii
Equilibrium Temp	200 – 320 Kelvin
Positional Noise	Total diff < 0.1

Each flag contributes 1 point to the habitability_score.

5. 🎯 KOI Disposition Classification

• Goal: Predict KOI disposition:
  CANDIDATE, CONFIRMED, FALSE POSITIVE, NOT DISPOSITIONED
• Features used:
  • Engineered features + habitability score
• Algorithms:
  • Random Forest Classifier
• Metrics:
  • Confusion Matrix
  • Classification Report
  • PCA Visualization

🧬 KOI Disposition Classes Explained

CANDIDATE: Probable planet, under review

CONFIRMED: Verified exoplanet

FALSE POSITIVE: Mistaken signal (e.g., stellar noise)

NOT DISPOSITIONED: Unclassified or unreviewed

---

Classification Report

              precision    recall  f1-score 

    CANDIDATE       0.35      0.19      0.25
    CONFIRMED       0.75      0.87      0.81
  FALSE POSITIVE    0.39      0.19      0.26 
 NOT DISPOSITIONED  0.70      0.68      0.69

📊 Visualisation

Features:

• Habitability Mapping (radius vs. temperature) /visual

• Confusion Matrix for model performance

• PCA plot for numeric feature space

---

🛠️ Tech Stack

Language: Python 3.x

Libraries:

Pandas, NumPy, Matplotlib, Seaborn

Scikit-learn (modeling + PCA)

---

📁 Project Structure

📦 exoplanet-habitability ┣ 📄 KeplerExoRaw.csv # Raw data

┣ 📄 cleankepler.csv # Cleaned dataset

┣ 📄 data.csv # Feature-engineered dataset

┣ 📄 featureEngineer.py # Feature engineering script

┣ 📄 model_train.py # ML model training & evaluatio

┣ 📄 README.md

---

DockerHub Upload

Link : https://hub.docker.com/repository/docker/newbieshubham/shubham-resume/tags/Exoplanet/sha256:ee4eb22206cdbd47dd5d7e0259cf203e1d6359a393863adfaebd05ca64cae75b

📌 Future Enhancements

Hyperparameter tuning for model

SHAP or feature importance analysis

Deployment via Docker or Streamlit Cloud

Integration with real-time exoplanet APIs

---

🙏 Acknowledgments

NASA Exoplanet Archive

Kepler Mission

---