Towards-History-aware-Sensitivity-Analysis-For-Time-Series

This repository contains the source code of our experiments.

Files

• decomposer.py: Class coding for the method proposed.
• make_experiments.py: Code to set up the synthetic experiment settings.
• run_experiments.py: Code to run synthetic experiments.
• load_data.py: Code to load real-world data.
• utils.py: Classes to generate synthetic data.
• run_real_exp.py: Code to run real data experiments.

Experiment Output

Running linear experiments will create the following folders:

• general_exp
• linear_short_exp
• linear_long_exp

Data Requirements

All necessary components to run synthetic experiments are available in the repository.

Real Data Experiments

To run real data experiments, you need to manually download the datasets from the following sources:

• GEFCom: GEFCom2012 Load Forecasting Data
• AirQuality: Beijing Multi-Site Air Quality Data

The CSV data files should be placed in the corresponding folders:

• data_gefcom
• data_airquality