Drug Category Classification

Introduction

This projects aims to build two deep neural network models, namely CNN and Bi-LSTM, to classify drug categories from a set of online drug reviews using Tensorflow/Keras and UCI Drug Review dataset.

Our models are able to classify 14 different drug categories from drug reviews with an overall accuracy of approximately 89%. An online version of this notebook is available on Colab.

How to run?

• Require python 3.8 or higher

• Install all dependencies for this project

$ pip install -r requirements.txt

• Start Jupyter Notebook

$ jupyter notebook

Experiment results

• CNN model with a Word Embedding layer using GloVe achives an accuracy of 89.24%

• Bi-LSTM model with a Word Embedding layer using GloVe achives an accuracy of 88.91%