This project aims to classify Amazon reviews into one of two categories, using several type of machine learning algorithms, including Logistic Regression and LSTM deep recurrent neural networks. The model is trained on the Amazon Review For Sentiment Analysis dataset. This dataset consists of 4,000,000 string with two labels - positive or negative.
There are several datasets publicly available for use in sentiment analysis. I decided to utilize the Amazon Review For Sentiment Analysis dataset.
- Python 3.8, Tensorflow 2.8, SKLearn, and NTLK.
- To install the required packages, run
pip install -r requirements.txt.
-
LSTM - 95.6% accuracy on validation set
-
MSM BERT - 94.2% accuracy on validation set
-
Logistic Regression - 90% accuracy on validation set
