THE СЕНСОРЛЫҚ ТЕХНОЛОГИЯЛАР МЕН МАШИНАЛЫҚ ОҚЫТУ: ЭКОЛОГИЯЛЫҚ МОНИТОРИНГТІҢ ЖАҢА МҮМКІНДІКТЕРІ

Abstract

This paper investigates the application of statistical machine learning methods for the calibration and analysis of data obtained from an artificial olfaction system during experimental studies with standard gas mixtures, which include key air pollutants (carbon dioxide, carbon monoxide, nitrogen dioxide, ammonia, and hydrogen sulfide) in concentrations ranging from 5 to 50 ppm. The relevance of the research is due to the need for accurate and reliable methods for monitoring air quality in conditions of low pollutant concentrations. Multivariate linear regression and clustering based on the nearest neighbors method were chosen as the machine learning methods. The resulting models demonstrated a high degree of agreement with the experimental data, as evidenced by the R2 coefficient values, which were close to one. Using the k-means algorithm, successful clustering of the multivariate sensor responses was carried out, revealing a clear relationship between sensor signal characteristics, gas type, and its concentration in the mixture. The research results can be used to create autonomous air quality monitoring systems capable of promptly detecting exceedances of the maximum allowable concentrations of harmful substances.