Система виявлення чадного газу на основі бездротового оптико-електронного сенсора концентрації з використанням машинного навчання

Abstract

The paper proposes a model using a random forest for a carbon monoxide detection system based on a wireless optoelectronic concentration sensor. In a certain environment that may be dangerous for humans, a sensor is placed that remotely transmits information about the gas concentration to a computer, where the information obtained is processed using machine learning. As a machine learning method, it is proposed to use the random forest method, since it provides high accuracy in conditions of heterogeneous data and complex nonlinearities, works well with minimal fine-tuning of hyperparameters, has high interpretability, stability and flexibility and is most suitable for carbon monoxide detection. The paper also proposes to use a random forest model for data processing of a wireless optoelectronic carbon monoxide concentration sensor. The study of the proposed carbon monoxide detection model based on a random forest was conducted through simulation, during which the accuracy of model training and validation, sensitivity, F-score, efficiency were evaluated, a confusion matrix was constructed by classes, and the importance of features was assessed.