BAYES CLASSIFIER

Bayes' Classifier is an optimal multiclass supervised classification method. Repository contains Python3 scripts for two-class classification on:

• toy-Gaussian data in 1D trained using MLE with Gaussian class conditional model,
• toy-Gaussian data in 2D trained using MLE with Gaussian class conditional model,
• toy-Gaussian data in 2D trained using MLE with Exponential class conditional model,
• toy-Gaussian data in 2D trained using EM with Gaussian Mixture Model,
• toy-Gaussian data in 20D trained using MLE with Gaussian distributed data,
• text corpus trained using naive Bayes' classifier with BoG model and TF-IDF features.

Documentation

docs/instructions.pdf contains the necessary instructions for the assignment, and docs/solutions.pdf contains the results and inferences.

Installation

Clone this repository and install the requirments using

https://github.com/kamath-abhijith/Bayes_Classifier
conda create --name <env> --file requirements.txt

Run

• /data/ contains the data files for the experiments. Description of the data are included in docs/solutions.pdf
• Run bayes_ex(x).py to run Bayes' classifier for exercise (x). Change the training_size and dataset variables.
• Run nn_ex(x).py to run nearest-neighbour classifier for exercise (x). Change the training_size and dataset variables.
• Run gmm_ex(x).py to run Bayes' classifier with Gaussian mixture model for exercise (x). Change the training_size and dataset variables.
• Run naiveBayes_doc_class.py to run sentiment analysis using naive Bayes' classifier.
• Find the results in results and the saved models in models.