COMP 472 - Assignment 1 (Winter 2021)

Team Members

• Michael Arabian
• Thomas Le
• Andre Saad

Introduction

For this assignment, we used Python along with the Scikit-learn machine learning framework to experiment with two different machine learning algorithms. We used a provided sentiment data set. The focus of this assignment was to gain experience on experimentations and analysis. See http://scikit-learn.org/stable/ for official documentation.

Instructions

• Download our GitHub repository as a Zip or use 'Git Clone' to have a copy on your computer.
  

git clone https://github.com/aramich100/COMP472_A1

• Make sure to have Python 3.9.2 installed on your computer. If you do not have it, you can install it here : https://www.python.org/downloads/
  
• Run the following command in your terminal in order to install scikit-learn and all necessairy libraries within the main.pu:

pip install scikit-learn

• To run the script, type the following command into the terminal. Make sure you are in the proper directory.

py main.py

• The program will now run the tasks in sequential order. Task 2 will display a plot and will pause the program. Once the plot from task 2 is closed, the remaining tasks will continue.

Our Results

As instructed, using the SciKit Framework, we were able to run 3 different Machine Learning Algorithms and obtained very promising results.

Naive Bayes

• Accuracy: 80.65463701216954
• Confusion Matrix: [ [ 1006 224 ][ 237 916 ]]

Decision Tree

• Accuracy: 72.2198908938313
• Confusion Matrix: [ [ 870 360 ] [ 302 851 ]]

Better Decision Tree

• Accuracy: 73.46454049517415
• Confusion Matrix: [ [ 868 362 ] [ 318 835 ]]

We can see that the Naive Bayes algorithm held the highest accuracy while compared to the Decision Tree. This is due to the fact that the Decision Tree is a discriminative model, whereas the Naive Bayes is a generative model. Given our data set, the Naive Bayes is best suited for the highest accuracy.