This is the Evaluation Tasks for the Google Summer Of Code 2021 Project Called DIMENSIONALITY REDUCTION FOR STUDYING DIFFUSE CIRCUMGALACTIC MEDIUM of The Organization Called Machine Learning for Science (ML4SCI) Umbrella Organization.
I was given a dataset that consists of 6 million labeled samples of two categories (classes) produced with Monte Carlo simulations.
Each sample consists of 28 features. The first 21 features are basic features related to the degrees of freedom of the problem.
The last seven are functions of the first 21.
Subtask 1: I have trained a Machine Learning Classifier such as
1)Logistic Regression,
2)KNN,
3)Decision Tree Classifier &
4)Random Forest classifier with ROC-AUC curve for each Classifier for the Classification of 2 Classes.
Subtask 2: I have used several Dimensionality Reduction Techniques Such as
1)Low Variation filter,
2)High Correlation filter,
3)Random forest feature importance,
4)Principal Component Analysis &
5)Truncated Singular Value Decomposition for dimensionality reduction of the dataset.
I also combined PCA & Truncated SVD with Logistic Regression with ROC-AUC Curve for comparing accuracy of original classifier and classifier with reduced dimensions.