KDDAutoML2019

This repository contains solution for AutoML in binary classification problems for temporal relational data. It was developed by a joint team ('autoflylearn') from Flytxt as a part of KDDCUP 2019 AutoML Challenge (The 5th AutoML Challenge: AutoML for Temporal Relational Data). Our solution improved significantly over the baseline solution provided by the organizers significantly and was one of the prominent solutions.

Team:

1. Harshvardhan Solanki (harshvardhan.solanki@flytxt.com)
2. Binay Gupta (binay.gupta@flytxt.com)
3. Amit Kumar Meher (amit.meher@flytxt.com)
4. Nasibullah Ohidullah(nasibullah104@gmail.com)

Contents:

• sample_code_submission/: Our solution code

How to run:

1. Install docker from https://docs.docker.com/get-started/.

2. Download starter-kit from competition organizer's website and replace their sample code solution with our folder "sample_code_submission"

3. At the shell, change to the startingkit directory, run

docker run -it --rm -u root -v $(pwd):/app/kddcup codalab/codalab-legacy:py3 bash

3. Now your are in the bash of the docker container, run ingestion program

cd /app/kddcup
python3 ingestion_program/ingestion.py

It runs sample_code_submission and the predictions will be in sample_predictions directory

4. Now run scoring program:

python3 scoring_program/score.py

It will score the predictions and the results will be in sample_scoring_output directory

Remark

• The full call of the ingestion program is:

python3 ingestion_program/ingestion.py local sample_data sample_predictions ingestion_program sample_code_submission

• The full call of the scoring program is:

python3 scoring_program/score.py local sample_predictions sample_ref sample_scoring_output