decode_reimplementation_outputs/README.md

DECODE Re-implementation

This folder is used by scripts/decode_reimplementation.py.

Goal

Re-implement the training flow from the DECODE paper for the local I-BLEND data:

1. Resample energy to 10-minute intervals.
2. Convert original power readings in W to interval energy in Wh.
3. Merge occupancy and calendar/schedule features.
4. Optionally merge weather if the local weather range overlaps the energy range.
5. Add historical energy features, including the previous three same-working-day-class values at the same time instant.
6. Normalize with Min-Max scaling.
7. Split chronologically into 70% train, 15% validation, and 15% test.
8. Train Ridge Regression, Decision Tree, Random Forest, and optionally LSTM. The LSTM backend prefers PyTorch when available and falls back to TensorFlow/Keras.

Install Dependencies

Base ML models:

python3 -m pip install pandas numpy scikit-learn

LSTM model, choose one backend:

python3 -m pip install torch
python3 -m pip install tensorflow

Run Examples

Train the 7 paper-style building targets:

python3 scripts/decode_reimplementation.py --mode paper_buildings

Train the 9 local meter targets:

python3 scripts/decode_reimplementation.py --mode meters

Quick smoke test on one target without LSTM:

python3 scripts/decode_reimplementation.py --mode paper_buildings --target Academic --max-rows 20000 --skip-lstm

Run one meter with LSTM:

python3 scripts/decode_reimplementation.py --mode meters --target Boys_main --epochs 20 --batch-size 64

Outputs

• processed/<target>_train_ready.csv: feature-engineered supervised table.
• results_paper_buildings.csv: model metrics for 7 building-level targets.
• results_meters.csv: model metrics for 9 meter-level targets.
• run_config_*.json: run metadata and paths.

Notes

The local weather file currently starts in 2018, while energy data ends in 2017. Because of this, the script does not force weather features into the model unless --include-weather is passed and an overlap is detected.