/ml-classification

Estimation and classification with traditional ML methods

Primary LanguageJupyter Notebook

ML Classifiers

  • Naive Bayes on the MNIST dataset. Classification accuracy = 84.03%
  • Perceptron on the IRIS dataset
  • Implementation and comparison b/w Maximum Likelihood and Expectation Maximization
  • Implementation and analysis of Bayes' classifier for high-dimensional data
  • Analysis of density estimation using Expectation maximization for unsupervised clustering
  • Understanding effect of different classification methodologies on synthetically generated data (sample from: Gaussian/Gamma/Uniform distributions)
  • Data exploration and analysis of the German Credit Score dataset for classification
  • Implementation of sub-sampling and oversampling strategies (in particular: SMOTE) for effective classification on the imbalanced German statlog dataset.
  • Analysis and verification of classifier performance on the IRIS dataset (ROC curves, decision boundaries, etc.).
  • Observing overfitting using polynomial regression by adjusting model complexity.