Compost Maturity and Emission Monitoring Dataset

{AUGMENTED} Compost Maturity Dataset from Khandakar et al., 2025

Dataset Description:

This dataset by Mullick and Mufti et al. was used in the paper titled "An Explainable and Scalable Framework for Compost Maturity Prediction Using Ensemble Learning and Conversational AI" published in ACM NSysS 2025 as a short paper, contains 1,314 compost-monitoring samples derived from an original set of 452 observations collected through an Arduino- and ESP32-based wireless sensor system. The data capture physicochemical dynamics of compost maturity and were expanded using SMOTE/SMOGN to balance minority ranges while preserving temporal and biological consistency. Each record corresponds to a composting instance within one of 94 batches and includes measured or derived maturity indicators widely used in environmental monitoring and waste-bioprocess analytics.

Features:

1. Day – Composting day at measurement time.
2. Temperature – Pile temperature (°C).
3. MC (%) – Moisture content.
4. pH – Acidity/alkalinity level.
5. C/N Ratio – Carbon–nitrogen ratio.
6. Ammonia (mg/kg) – Ammoniacal nitrogen concentration.
7. Nitrate (mg/kg) – Nitrate nitrogen concentration.
8. TN (%) – Total nitrogen.
9. TOC (%) – Total organic carbon.
10. EC (mS/cm) – Electrical conductivity.
11. OM (%) – Organic matter content.
12. T Value – Transformation value representing biodegradation progression.
13. GI (%) – Germination Index, a phytotoxicity indicator.
14. Score – Compost maturity score.
15. Batch – Assigned compost batch (1–94).
16. Synthetic – Binary flag indicating SMOTE-generated samples.
17. Source – Original or augmented label.

Distribution Summaries:

All numerical variables include binned label–count summaries to support exploratory data analysis, outlier screening, and range-based modelling strategies. These distributions reveal wide variance in nitrogen compounds, organic matter decomposition rates, and moisture–temperature transitions across the composting timeline.

Sampling Notes:

• 34% of samples originate from the real sensor-acquired dataset.
• 66% are SMOTE/SMOGN-generated to enhance representation across maturity stages.
• Time-continuous and batch-based patterns are preserved to enable forecasting, classification, and explainability analyses.

The original dataset (452 samples) is available from Khandakar et al. (2025):

https://www.sciencedirect.com/science/article/abs/pii/S0045790625000588

Citation for This Augmented Dataset (1,314 samples):

If you use this dataset, please cite:

@inproceedings{MullickandMufti2025,
  title={An Explainable and Scalable Framework for Compost Maturity Prediction Using Ensemble Learning and Conversational AI},
  author={Mullick, Mohammad Abu Obaida and Mufti, Abu Henaf Rashid Ahmad and Ratul, Rezaur Rahman and Noor, Jannatul},
  booktitle={Proceedings of the 12th International Conference on Next Generation Computing, Communication, Systems and Security},
  pages={29--34},
  year={2025},
  url = {https://doi.org/10.1145/3777555.3777574},
  doi = {10.1145/3777555.3777574}
}