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.
-
Require
python 3.8or higher -
Install all dependencies for this project
$ pip install -r requirements.txt
- Start Jupyter Notebook
$ jupyter notebook
- 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%
