Yelp-Recruiting-Competition

Data Mining G07 Final Project (Kaggle competition link: https://www.kaggle.com/c/yelp-recruiting)

Requirements

numpy == 1.14.3
pandas == 0.23.4
sklearn == 0.20.0
keras == 2.1.6
tensorflow == 1.9.0

Dataset

Please first download the dataset from Kaggle to the directory 'data/'. (https://www.kaggle.com/c/yelp-recruiting/data)

Directories and files

$ tree
.
├── data (need to download first)
│   ├── yelp_test_set
│   │   ├── yelp_test_set_business.json
│   │   ├── yelp_test_set_checkin.json
│   │   ├── yelp_test_set_review.json
│   │   └── yelp_test_set_user.json
│   ├── yelp_training_set
│   │   ├── yelp_training_set_business.json
│   │   ├── yelp_training_set_checkin.json
│   │   ├── yelp_training_set_review.json
│   │   └── yelp_training_set_user.json
│   └── sample_submission.csv
├── feature
│   ├── test_X.pkl
│   ├── X.pkl
│   └── Y.pkl
├── model
│   └── DNN.h5
├── submit
│   ├── DNN.csv
│   ├── ensemble.csv
│   ├── LR.csv
│   ├── RFR.csv
│   └── SVR.csv
├── data_preprocessing.py
├── DNN.py
├── ensemble.py
├── LR.py
├── README.md
├── Report.pdf
├── RFR.py
└── SVR.py

6 directories, 26 files

Data Preprocessing

After downloading the dataset, run:

$ python3 data_preprocessing.py

Check the saved file 'test_X.pkl', 'X.pkl', 'Y.pkl' under the directory 'feature/'.

Prediction

Using Linear Regression (LR), run:

$ python3 LR.py

Check the saved file 'LR.csv' under the directory 'submit/'.

Using Deep Neural Network (DNN), run:

$ python3 DNN.py

Check the saved file 'DNN.csv' under the directory 'submit/'.

Using Support Vector Regressor (SVR), run:

$ python3 SVR.py

Check the saved file 'SVR.csv' under the directory 'submit/'.

Using Random Forest Regressor (RFR), run:

$ python3 RFR.py

Check the saved file 'RFR.csv' under the directory 'submit/'.

Before using ensemble, please first check that 'SVR.csv', 'DNN.csv' and 'RFR.csv' are under the directory 'submit/'. Then run:

$ python3 ensemble.py

Check the saved file 'ensemble.csv' under the directory 'submit/'.

Evaluation

Method	Private Score	Public Score	Ranking
LR	0.50826	0.50889	94/350
DNN	0.48545	0.48405	62/350
SVR	0.50575	0.50356	87/350
RFR	0.52554	0.52579	120/350
ensemble	0.48340	0.48241	59/350

Evaluation Metric: Root Mean Squared Logarithmic Error ("RMSLE")