Machine Learning TA Helper

A simple toolkit for accessing data from Kaggle and post-process for calculating student grades and finding potential cheatings 📝🎓.

Requirements

Python 3

No extra packages needed.

Installation

git clone https://github.com/cyhuang-tw/ml-ta-helper.git

Usage

Download Kaggle Leaderboards

• Public Leaderboard:
  Directly dowload the .csv file from Kaggle.
• Private Leaderboard:
  Download the whole webpage as a plain text file.

Calculate Kaggle Scores

<student list>.csv is required, and the columns should be in the order of index, student ID, orginal student ID.

python3 collect_score.py \
    --public <public leaderboard>.csv \
    --private <private leaderboard>.html \
    --students <student list>.csv \
    --publ-bl <public baselines (from easy to hard)> \
    --priv-bl <private baselines (from easy to hard)> \
    --output <final output file>.csv

Download Kaggle Submission Records

python3 get_kaggle.py \
    --competition_id <competition ID>

This step might take some time if there are thousands of teams in the competition.
The downloaded .json files recording the submission of the students are in the directory kaggle_output/single_student/ while the team ID to team name mapping will be saved in kaggle_output/id2name.json.

Find Potential Cheatings

python3 find_cheating.py \
    --dir kaggle_output/single_student \
    --id2n kaggle_output/id2name.json \
    --output <output file>.txt \
    [--descending]

The potential cheatings will be stored in a .txt file.

Convert to NTU COOL-compatible Files

Download a fresh copy of the score file from NTU COOL.
Download a .csv file containing the NTU COOL ID and score of each student.

python3 convert_to_ntucool.py \
    --orig-file <original score file (w/ ID)>.csv \
    --id-col <column index of ID> \
    --score-col <column index of scores> \
    --cool-grade <original NTU COOL grade file>.csv \
    --cool-output <output NTU COOL grade file>.csv \
    --title "<title for the new added HW>"