predicting yelp's elite

A project to predict a Yelp user's elite status based on user activity, popularity/social network size, review sentiment/content, and review structure. Investigates various classification models, including random forest classifiers, naive bayes, logistic regression, and SVM. Final model achieves an accuracy score of 98% and AUC of 99%.

For more information, see my blog post.

in this repo

• yelp-text-processing.ipynb jupyter notebook with scripts and outputs for processing review text
• yelp-classification.ipynb jupyter notebook with scripts and outputs for feature engineering and model selection
• d3/ contains scripts for feature importance bar chart and network graph viz
• presentation/ contains pdf presentation of findings & recommendations

For the sake of not overloading github, the original data was not uploaded to this repo. Yelp data can be accessed at the link provided at the bottom of this readme.

installation

clone this repo

$ git clone https://github.com/dianalam/yelp-classifier.git

dependencies

Scripts were written in Python 2.7. You'll need the following modules:

matplotlib >= 1.5.1  
nltk >= 3.1
numpy >= 1.10.1  
pandas >= 0.17.1  
python-dateutil >= 2.4.2
scipy >= 0.16.0
seaborn >= 0.6.0
sklearn >= 0.17
spacy >= 0.100
statsmodels >= 0.6.1

To install modules, run:

$ pip install <module>

running

To open jupyter notebooks:

jupyter notebook 

To run visualizations:

python -m SimpleHTTPServer

Then navigate to instantiated port.

data sources

Thanks to:

• Yelp Dataset Challenge