Selective Sensing of Volatile Organic Compounds Using\nan Electrostatically Formed Nanowire Sensor Based on Automatic Machine\nLearning

Abstract

With\nthe development of Internet of Things technology, various\nsensors are under intense development. Electrostatically formed nanowire\n(EFN) gas sensors are multigate Si sensors based on CMOS technology\nand have the unique advantages of ultralow power consumption and very\nlarge-scale integration (VLSI) compatibility for mass production.\nIn order to achieve selectivity, machine learning is required to accurately\nidentify the detected gas. In this work, we introduce automatic learning\ntechnology, by which the common algorithms are sorted and applied\nto the EFN gas sensor. The advantages and disadvantages of the top\nfour tree-based model algorithms are discussed, and the unilateral\ntraining models are ensembled to further improve the accuracy of the\nalgorithm. The analyses of two groups of experiments show that the\nCatBoost algorithm has the highest evaluation index. In addition,\nthe feature importance of the classification is analyzed from the\nphysical meaning of electrostatically formed nanowire dimensions,\npaving the way for model fusion and mechanism exploration.