E3-JSI / iot-fusion

Streaming data fusion framework adjusted to support real-world use cases in the IoT domain.

IoT-fusion component

IoT fusion is a streaming data fusion system that can create machine-learning-ready feature vectors out of various heterogeneous sensor streams (i.e. sensor data, weather forecasts and static data). The system consists of the following components:

• fusion (stream fusion component with integrated incremental learning models)
• modeling (separate modeling component based on Python's scikit-learn models, LightGBM and xgboost)
• server (administrative server for communication with distributed fusion components)
• client (GUI for interaction with server nad further with fusion components)

Docs

Documentation is available here.

Publications

In case you use any of the components for your research, please refer to (and cite) the paper:

• Kenda, K.; Kažič, B.; Novak, E.; Mladenić, D. Streaming Data Fusion for the Internet of Things. Sensors 2019, 19, 1955.

Applications

• Kenda, K.; Mellios, N.; Senožetnik, M.; Pergar, P. Architecture for Stream Mining in Water Management. (in preparation)

Related work

• Kenda, K.; Kažič, B.; Stopar, L., Fortuna, B., Rupnik, J., Škrjanc, M.; Mladenić, D. Data Fusion Framework for Streaming Heterogeneous Data Sources (in preparation)
• Kenda, K., & Mladenić, D. Autonomous sensor data cleaning in stream mining setting. Business Systems Research Journal, 9(2), 69-79.