Prediction Star Review From Its Text Alone

Capstone Project for Coursera Data Science Specialization

• For this project we train 1.6 million reviews,total Yelp Dataset.
  
• Main steps of the project:
  

• We used-LDA for sentiment analysis in star review.

1.Detecting the 1, 3 and 4-6 words of users text input.
2.Each detected words have rating system based on extracted star.
3.Calculating mean of the detected words rates, which it corresponds with star rate.

-Predicted accuracy for matched predicted test star and real star is: 58%.
-Predicted accuracy for approximation test star and real star is: 88%.
Real star is 5, predicted star should be 5 or 4 but not 1, 2 and 3…