Smart Breath Analysis for IoMT Consumer Electronics: Feature Engineering for Health Applications

Smart breath analysis is rapidly emerging as a cornerstone in the development of internet of medical things (IoMT) consumer electronics for non-invasive, real-time health monitoring. However, effective deployment in holographic health applications demands robust feature engineering techniques capable of handling the challenges posed by sensor cross-sensitivity. One such challenge is correlation bias (CB), where highly correlated features reduce the effectiveness of traditional selection algorithms, such as twin support vector machine recursive feature elimination (TWSVM-RFE), by obscuring the importance of truly discriminative features. In this paper, we address this critical issue by introducing a novel feature grouping method, infinite correlation grouping (ICG), designed to mitigate CB in the context of IoMT-driven breath analysis systems. ICG constructs an undirected, fully-connected weighted graph to cluster features into highly correlated groups and selects group centers as representative in the recursive elimination process. Extensive evaluation on both synthetic and real-world breath datasets shows that our method not only alleviates the impact of CB but also significantly enhances classification performance. Meanwhile, the additional time complexity introduced by the ICG relative to the baseline feature selection algorithm is only O(n*d2+d3), thereby ensuring the feasibility of the proposed method on time-sensitive IoMT devices. These findings underline the potential of ICG as a foundational tool for smart feature engineering in next-generation holographic health technologies.