Predicting the Useful Values with Machine Learning Models
1. Figure
[Fig. Random forest features by importance]
[Fig. Result of models]
2. Goal
To determine which machine learning model provides the best prediction of whether a review on California restaurants on Yelp will be voted as “useful” by users.
3. Methodology & Summary
- Random Forest has an AUC score of 91.58% that shows the model is learning the data well enough.
- By using the Random Forest model that we made, the restaurants can predict whether a review is useful or not even before people start pressing a useful button on Yelp. If a review is useful, which is written as a reasonable good, bad or both things, the restaurants need to change the customers’ complaints or maintain the good contents the reviewers have told for their restaurants’ development.