Solution for Elisa AI Challenge 2018

For more information about the challenge and data see https://www.elisa.ee/et/elisa-ai-challenge and https://gitlab.com/TeamMindTitan/workshop-naai

Dependencies

Requires Python 3.6

This project uses Pipfile to manage its Python dependencies.

Model Training

The model training can be performed using script train_model.py. Training data file paths or urls (normals.parquet and 'probs.parquet') must be provided as command line arguments. Optionally these can be provided as an environment variable TRAIN_FILES as comma separated list. The resulting model will be saved under models directory together with training log.

$ pipenv run python scripts/train_model.py data_file [data_file]...

Inference

The inference script is run_model.py.

Both the model file path and data file path must be provided as command line arguments.

If the model file is omitted, the latest model file from models directory will be chosen. Optionally the data file can be provided as an environment variable RUN_FILES as comma separated list.

$ pipenv run python scripts/run_model.py [model_file] data_file

The results will be saved under models directory combining under as csv file basename(model_file)-basename(data_file).csv