Product Rating Analysis

Inspiration

Creating a prototype that conducts sentiment analysis on reviews to improve products and customer service.

What it does

• Utilizes ML to identify positive, negative and neutral sentiments in reviews.
• Visualizes which products had most positive or negative reviews, the frequency of positive and negative words, average rating by product, and the number of reviews per product.

How we built it

• Python for modelling sentiment

  • Jupyter Notebook
  • Pytorch
  • NumPy
  • Pandas

• Tableau for visualization

Challenges we ran into

• Finding a suitable dataset to train the model and use for a mock analysis
• Achieving sufficient accuracy on the sentiment analysis model
• Using PyTorch correctly to build and train a converging text classification model

Accomplishments that we're proud of

• We were able to pass datasets through the ML-based classifier that identifies different sentiments. We achieved an accuracy of 63% in a 24 hour time-frame, starting from an accuracy of 34%. For comparison, sentiment analysis models are generally considered to be nearly as good as a human if it has 70% accuracy.
• We then analyzed the data by product and visualized the results, so that we could provide beneficial suggestions to improve the customer experience.

Built With

• python
• pytorch
• tableau