Machine Learning: Exoplanet Exploration

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, I created machine learning models capable of classifying candidate exoplanets from the raw dataset. The steps taken to do so are outlined below:

Directory:

• machine-learning-challenge (Repo Home):
  • README.md
  • exoplanet_data.csv ---> Source: Kepler Exoplanet Search Results
  • exoplanet_exploration_randomForest.ipynb
  • exoplanet_exploration_svm.ipynb
  • Images (Folder):
    • featureimportance.png
    • randomforest_classificationreport.png
  • Models (Folder):
    • zGrinacoff_randomForest.sav
    • zGrinacoff_svm.sav

PLEASE USE RANDOM FOREST MODEL

Project Outline:

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 two different classifiers (Support Vector Machine & Random Forest Classifier).

Report and Final Model Outcome:

• Even without controlling GridSearchCV params, the Random Forest Model was much more effective and capable in classifying candidate exoplanets.

• Random forests are much simpler to train for a practitioner; it’s easier to find a good, robust model. The complexity of a random forest grows with the number of trees in the forest, and the number of training samples we have. In SVMs, we typically need to do a fair amount of parameter tuning, and in addition to that, the computational cost grows linearly with the number of classes as well.

Random Forest:

	BeforeCV	AfterCV
Training Score	0.995	1.0
Testing Score	0.883	0.899

Random Forest Classification Report:

Random Forest Feature Importance:

Support Vector Machine:

	BeforeCV	AfterCV
Training Score	0.846	0.887
Testing Score	0.842	0.879

• Although the Random Forest model was quite successful, it would be important to train and test additional models for accuracy and reliability. However, for a first time Machine Learning project, I am quite happy with the results. I look forward to further exploration in Machine Learning and beyond our planet's solar system!