Evolving Neuro-Fuzzy Systems for Real-Time Data Processing in Internet of Medical Things

Abstract Evolving neuro-fuzzy systems (ENFS) have shown great promise in analysis of streaming data, integrating artificial neural networks and fuzzy logic to create self-learning, data-driven models suitable for online (real-time) applications. In healthcare, ENFS have the potential to transform the Internet of Medical Things (IoMT) by enabling real-time data processing, decision support, and timely emergency response. In this study six evolving fuzzy min-max neural network models have been implemented using Python and Kafka, and evaluated on the MIT-BIH Arrhythmia dataset comprising 109,446 samples corresponding to shapes of heartbeats for the normal case and the cases affected by different arrhythmias and myocardial infarction. To achieve online continual learning, the prequential test-then-train method has been used to sequentially classify the data instances from scratch and update the system structure along with the arrival of new data samples. The Enhanced Fuzzy Min-Max Neural Network (EFMMNN) achieved the highest accuracy of 98.11%. Future work will explore deploying these evolving fuzzy systems in real-world healthcare settings using edge-fog computing.