Skip to content

Understanding Exoplanets with Data Science is about testing efficiency of Machine learning Classifiers on highly imbalanced data such as Kepler-planet-catalog and Nasa’s Exoplanet Archive

Notifications You must be signed in to change notification settings

prkhrv/Understanding-Exoplanets-with-Data-Science

Repository files navigation

Understanding-Exoplanets-with-Data-Science

Understanding Exoplanets with Data Science is about testing efficiency of Machine learning Classifiers on highly imbalanced data such as Kepler-planet-catalog and Nasa’s Exoplanet Archive. The Project consists of two parts

  1. Predicting the presence of Exoplanets
  2. Predicting the Habitability of confirmed Exoplanets

Part 1: - Predicting the presence of Exoplanets

For this part, I have used Kepler’s Cumulative Planet archive -2021 which contains all the potential and confirmed planets with their parameters. After filtering data for missing values and preprocessing we test our dataset for prediction in several Classifiers such as

  • Bagging Classifier
  • Decision Trees
  • Random Forest
  • Logistic Regression
  • KNN
  • Naïve Bayesian

We get the Following Results: -

result1

Part 2: - Predicting the Habitability of confirmed Exoplanets

To Predict the habitability of exoplanets we have merged two datasets which are 1. Nasa’s Exoplanet Archive 2. PHL (Planet Habitability Laboratory) Catalogue. To fix missing data I have used Imperative Imputation and after that I have tested the data against the various Classifiers. Here is the accuracy result: -

result2

Results: -

For Planet Detection Confirmation Bagging Classifier/ Decision Trees works the best For Habitability Prediction XGBoost is the best Classifier.

How to Run:

The project only requires some packages which are used for analysis and visualization

  • In the project Folder run: pip install -r requirements.txt
  • It will install all the dependencies required for this project
  • All the data files are separately stored in “data” folder for each part.

About

Understanding Exoplanets with Data Science is about testing efficiency of Machine learning Classifiers on highly imbalanced data such as Kepler-planet-catalog and Nasa’s Exoplanet Archive

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published