- Dataset : 400 thousand reviews of unlocked mobile phones sold on Amazon.com
- Problem : Sentiment analysis for Amazon product reviews
- Use Natural Language Procesisng techniques, Bag of Words model, Word2Vec model and Long Short Term Memory (LSTM) neural network to conduct sentiment analysis for Amazon product reviews.
- Accuracy score of 94.4% by Word2Vec embedding with LSTM.
amazon-sentiment-analysis.ipynb
- Data visualization of Amazon product reviews
- Preprocess raw reviews to cleaned reviews
- Use different word embedding models, such as count vectorizer, tf-idf transformation and Word2Vec model, to transform text reviews into numerical representations
- Fit numerical representations of text reviews to LSTM (a type of recurrent neural network)
For more detailed information, see Report.