Machine_Learning_Exoplanet_Exploration

Processing NASA Kepler space telescope data through machine learning models capable of classifying candidate exoplanets from the raw dataset.

Background

Over a period of nine years in deep space, the NASA Kepler space telescope has been out on a planet-hunting mission to discover hidden planets outside of our solar system.

To help process this data, you will create machine learning models capable of classifying candidate exoplanets from the raw dataset.

In this project, you will need to:

1. Preprocess the raw data
2. Tune the models
3. Compare two or more models

---

Instructions

Preprocess the Data

• Preprocess the dataset prior to fitting the model.
• Perform feature selection and remove unnecessary features.
• Use MinMaxScaler to scale the numerical data.
• Separate the data into training and testing data.

Tune Model Parameters

• Use GridSearch to tune model parameters.
• Tune and compare at least two different classifiers.

Reporting

• Random Forest model is better than Support Vector Machine, even without hyper tuning the parameters with GridSearchCV. Of all the models considered (SVM, Random Forest and Gradient Boosting). Gradient Boosting model yielded highest accuracy of 91%.

SVC

                precision    recall  f1-score   support

     CANDIDATE       0.76      0.61      0.68       422
     CONFIRMED       0.69      0.79      0.74       450
FALSE POSITIVE       0.99      1.00      0.99       876

      accuracy                           0.85      1748
     macro avg       0.81      0.80      0.80      1748
  weighted avg       0.85      0.85      0.85      1748

SVC with Hyperparameter Tuning

                precision    recall  f1-score   support

     CANDIDATE       0.84      0.71      0.77       422
     CONFIRMED       0.76      0.86      0.80       450
FALSE POSITIVE       0.99      1.00      0.99       876

      accuracy                           0.89      1748
     macro avg       0.86      0.85      0.86      1748
  weighted avg       0.89      0.89      0.89      1748
  

Random Forest

              precision    recall  f1-score   support

   CANDIDATE       0.76      0.84      0.80       378
   CONFIRMED       0.85      0.82      0.84       469
FALSE POSITIVE     1.00      0.97      0.98       901

    accuracy                           0.90      1748
   macro avg       0.87      0.88      0.87      1748
weighted avg       0.91      0.90      0.90      1748

Random Forest with Hyperparameter Tuning

              precision    recall  f1-score   support

   CANDIDATE       0.84      0.75      0.79       422
   CONFIRMED       0.81      0.86      0.83       450
FALSE POSITIVE     0.97      1.00      0.99       876

    accuracy                           0.90      1748
   macro avg       0.88      0.87      0.87      1748
weighted avg       0.90      0.90      0.90      1748

Gradient Boosting

                precision    recall  f1-score   support

   CANDIDATE       0.85      0.79      0.82       422
   CONFIRMED       0.82      0.85      0.83       450
FALSE POSITIVE     0.99      1.00      0.99       876

    accuracy                           0.91      1748
   macro avg       0.88      0.88      0.88      1748
weighted avg       0.91      0.91      0.91      1748

Gradient Boosting with Hyperparameter Tuning

              precision    recall  f1-score   support

   CANDIDATE       0.84      0.80      0.82       422
   CONFIRMED       0.82      0.85      0.83       450
FALSE POSITIVE     0.99      1.00      1.00       876

    accuracy                           0.91      1748
   macro avg       0.88      0.88      0.88      1748
weighted avg       0.91      0.91      0.91      1748

---

Resources

• Exoplanet Data Source

• Scikit-Learn Tutorial Part 1

• Scikit-Learn Tutorial Part 2

• Grid Search

---