Exoplanet Classification Using Machine Learning

Exploring distant worlds by combining astronomy with machine learning.

Comment

Exoplanet Classification Using Machine Learning

Inspiration

I have always been fascinated by astronomy, especially the search for planets beyond our solar system. When I discovered the India High School Exoplanet Data Challenge, I saw it as an opportunity to combine my interest in space with machine learning. I wanted to explore how AI can help scientists analyze large astronomical datasets and identify potential exoplanets more efficiently.

What it does

This project classifies objects from NASA's Kepler Objects of Interest (KOI) dataset into three categories:

  • Confirmed
  • Candidate
  • False Positive

The model helps automate the classification process, making it easier to analyze thousands of observations.

How I built it

I developed the project in Python using Jupyter Notebook.

The workflow included:

  • Loading the NASA KOI dataset using Pandas.
  • Exploring and understanding the data.
  • Removing unnecessary columns.
  • Handling missing values.
  • Encoding categorical features.
  • Splitting the dataset into training and testing sets.
  • Training a Random Forest Classifier.
  • Evaluating the model using accuracy, precision, recall, F1-score, and a confusion matrix.
  • Visualizing feature importance.

The final model achieved an accuracy of approximately 90.7%.

Challenges we ran into

This was my first machine learning project using a real-world scientific dataset.

One challenge was learning how to use Jupyter Notebook and organize an ML workflow. Another challenge was dealing with missing values and deciding which features to keep or remove before training the model. Understanding the preprocessing steps was just as important as building the model itself. But the biggest challenge was the limited time available. I discovered the competition only five days before the deadline and had no prior experience using Jupyter Notebook. This required me to quickly learn the workflow, understand the dataset, build the model, and prepare the final submission within approximately 50 hours.

Accomplishments that we're proud of

  • Built my first complete machine learning project.
  • Worked with a real NASA dataset.
  • Achieved approximately 90.7% accuracy.
  • Created visualizations to understand the model.
  • Published the project on GitHub.

What I learned

Through this project I learned how to:

  • Explore and clean datasets.
  • Preprocess data for machine learning.
  • Train and evaluate a classification model.
  • Interpret metrics such as precision, recall, F1-score, and accuracy.
  • Use GitHub to share projects.

Most importantly, I learned how machine learning can be applied to real scientific problems like exoplanet discovery.

What's next for Exoplanet Classification Using Machine Learning

In the future, I would like to:

  • Compare multiple machine learning algorithms.
  • Perform hyperparameter tuning.
  • Use cross-validation.
  • Try advanced models such as XGBoost.
  • Explore additional feature engineering techniques to improve accuracy.

This project is only the beginning of my journey into astronomy and machine learning.

Built With

Share this project:

Updates

Submission history