This project analyzes and develop a model to analyze and predict the sentiment given the costomer review for an amazon baby product. Turicreate is used for this work. The data set is highly biased towards the positive sentiment, so model is improved, by using only selected features for training the model.
Rating for the most reviewed product
Product rating analysis as general
Sentiment Analysis
0 - 1 is for negative sentiment, and 1-2 is for positive sentiment
Developed model uses Logistic Regression algorithm. Performance metrics:
- 'f1_score': 0.9157860082304526,
- 'log_loss': 0.39622654670874996,
- 'precision': 0.8487520595594068,
- 'recall': 0.9943165570488991
For code: Notebook