CS760

This repositories is for CS760 Group 6. Welcome guys!

Our research topic is:

Predicting Online Yelp Review Usefulness

Scripts

To run anything in scripts/, run from the project root directory:

bash ./scripts/myScript

Script Descriptions:

• exportConda & importConda: export or import the conda environment via environment.yml. Used for recreating a consistent conda environment.
• startJupyter: Starts a jupyter lab server. Open jupyter in your browser by clicking on the URL printed out on the terminal.
• initDB: Create a new database file with empty tables (see database/scripts/createTables.sql). Deletes old database file if exists.

Raw Data

Please place raw data (straight from the internet) into preprocessing/raw_data/. This is so that we have a consistent location for our code to operate on.

A Note on Random Seeds

Please set random seeds to some constant so that we can keep multiple runs consistent.

Conda Environment

For reproducibility and standard experimentation environment, a conda environment should be used. The conda environment can be built from environment.yml by running the script ./scripts/importConda. The libraries and tools included are those in CS 762 (numpy, scikit-learn, pandas, jupyterlab, matplotlib). More can be added as needed.

Make sure to activate the environment by running:

conda activate CS760

To deactivate:

conda deactivate

Preprocessing

• Importing raw data & saving to disk: Run initDB (see scripts) first. Then run importYelp.ipynb to import data into SQL, join relevant columns, and save to disk as a compressed file preprocessing/processed_data/joined.parquet.snappy
• Note: this requires fastparquet package. Should already be installed if you duplicated the conda environment.