Machine learning algorithm assisted cerium oxide based high selectivity acetone sensor

Accurate and quick determination of acetone concentration is crucial for the diagnosis and prevention of health risks. However, semiconductor metal oxide-based gas sensors have poor selectivity for organic volatile gases. In this study, the machine learning algorithm is employed to improve the performance of acetone sensors based on indium-doped cerium dioxide. CeO 2 –In series sensors are prepared by adjusting the doping ratio of indium in CeO 2 nanorods. CeO 2 –In10 % to 100 ppm acetone show a response (2.52) of 2.17 times higher than pure CeO 2 at 300 °C. Indium doping increases specific surface area and reduces response time to 1.1 s. Based on the obtained gas-sensitive data, a machine learning algorithm is used to analyze the gas types. Machine learning classifies gases with 97.5 % accuracy (Random Forest). The idea of using machine learning to improve the selectivity of acetone sensors may provide a new way for the study of metal oxide semiconductor gas sensors with high selectivity.