Skip to main content

Twitter Sentiment Analysis Classifier

A project for CPTS 437 - Introduction to Machine Learning at WSU, where we implemented multiple traditional machine learning algorithms (along with a simple Neural Network) from scratch to classify tweets based on their sentiment.

Algorithms Tested:

  • K Nearest Neighbor (using embedding model to convert tweet to vector space)
  • Support Vector Machine
  • Naive Bayes Predictor
  • Neural Network

In the end, our Neural Network achieved the highest accuracy (~70%), which we are happy with due to implementing that algorithm ourselves. Next was SVM at 62% (but much slower 😢), KNN at 55%, and Naive Bayes at ~45% to ~50% (due to data processing issues)

Aidan Johnson
Author
Aidan Johnson
A Master of Computer Science student at Washington State Univeristy, studying software design principles and object-oriented programming.