sanyuzhang / TripAdvisorReviewRegressor

A regression model based on LightGBM to predict the rating by parsing the review using NLTK

TripAdvisorReviewRegressor

A LightGBM model based regressor which can predict the overall rating of a hotel by parsing its reviews using NLTK

Env Setup

1. Using Python3. All the commands are using python3.

2. If you have both python2 and python3 env, do:

   • Replace python to python3
   • Replace pip to pip3

3. Install all the packages listed in the requirements.txt

   • pip install -r requirements.txt

How to Run

1. Type the command below.
   • python reviews_classifier.py
     • Train the processed data with features extracted, which takes about 10+ seconds.
   • python reviews_classifier.py full
     • Run the full code including feature extractions using NLTK, which takes 10+ minutes.

Folders

1. All the raw data are under the data/
2. processed_data.csv are data of features extracted from raw data using NLTK.
3. prediction.csv are a table of true ratings and our predicted ratings.