Ridge Regression-based Approach for Humidity Estimation in Differential Mobility Spectrometry

Abstract

Analyzing differential mobility spectrometry (DMS) data recorded in uncontrolled environmental conditions is challenging. Understanding environmental effects on the spectra opens a new path for standardizing measurement results before the subsequent automated analysis with machine learning algorithms. While several parameters influence the spectra, humidity stands out as the most dominant factor, making it the focus of this analysis. In this work we applied two feature engineering methods and implemented the ridge regression algorithm on DMS data to estimate humidity level. The ridge regression can accurately estimate humidity levels for given DMS spectra. Furthermore, the analysis confirmed that the most critical information for processing DMS spectra is contained in the shape of the dispersion curves.