Machine learning-driven gas concentration forecasting: A case study with WS 2 nanoflower gas sensor

One approach to effectively forecast gas concentrations is to integrate machine learning techniques with gas-sensitive signals. In this work, the synthesis of two-dimensional nanoflower-like WS 2 was achieved through control of reaction conditions, such as hydrothermal time and temperature, accompanied by an analysis of the formation mechanism underlying the nanoflower-like morphology. At room temperature, the WS 2 nanoflower exhibited a notable response value of 51.61 % towards 100 ppm NH 3. Subsequently, variations in the response of WS 2 to NH 3 were examined under diverse conditions. Two methodologies were employed for parameter extraction and transient signal analysis to construct the eigenvector. Accurate prediction of NH 3 concentration was achieved using four machine learning models, namely ANN, DT, LR and RF. Notably, the RF model exhibited prediction accuracy of 92 % and 84 % for two distinct vectors. By employing parameter, the accurate forecasting of NH 3 concentration was facilitated, broadening the temperature range applicable to the WS 2 gas sensor.