Paper - Tourism Forecast with Weather, Event, and Cross-industry Data
Workflow to train models
Build docker for debugging and building
docker build -t tourism_docker -f ./docker/Dockerfile-dev ./
Call the train function
docker run -v "$(pwd)"/data:/app/data -v "$(pwd)/results":/app/results -v "$(pwd)/tourism_workflow":/app/main -it tourism_docker /bin/bash -c "cd /app/main && python -m scripts.train --help"
Call nvidia-docker for gpu 1
docker run --gpus "device=1" -v "$(pwd)"/data:/app/data -v "$(pwd)/results":/app/results -v "$(pwd)/tourism_workflow":/app/main -it tapfaeff-tourism-gpu-stack
Call if using the sources as python library.
python -m scripts.preprocess_dataset --base_path /data --dataset dataset --num_train_years 4 -v
To plot the target value for a dataset use
python -m plotting.plot_data --dataset_path /app/data/dataset/dataset.csv --dataset_delimiter ',' --column_information_path /app/data/dataset/dataset.csv --column_information_delimiter ";" --save ../dataset_plot.png