Metal oxide (MOX) gas sensor is one of the technology that has been widely used in different applications. For decades MOX gas sensor has been used for gas monitoring due to its advantages, such as high sensitivity, high detection range, fast reaction time and cost-effectiveness. However, MOX gas sensor has limitation in several factors that affect the sensing ability, such as chemical composition, operating temperature, ambient temperature and humidity. In this research, the cross-sensitivity of the gas sensors toward ambient temperature and humidity. PCB boards were developed, which consists of temperature and humidity sensors, as well as eight different MOX gas sensors. Also, a partially closed chamber was fabricated to allow inflow and outflow of air for the ambient temperature and humidity control. The sensors were subjected to various temperatures (from 16˚C to 30˚C), humidity (from 75% to 45%), which are relevant to typical indoor environment. At each of these parameter settings, the gas sensor responses were continuously recorded at different ethanol gas concentrations (i.e. 0%, 0.05%, 0.2%, 0.5%,1% and 2%). Based on the results, it was proven that the gas sensor responses are affected by the temperature and humidity. The increase of temperature and humidity levels lead to the decreased of gas sensor response for most of the sensor, except for MiCS-6814 (NH3 sensor) which showed the opposite response. Linear regression based on machine learning approach was applied to correct the gas sensor response drift by producing several models. The models were validated by the technique of K-fold Cross Validation (CV) to provide the measure of fit to the candidate models with respect to the experimental data. Finally, the best fit model was verified and proven to be able to minimize the sensor response drift due to temperature and humidity for all gas sensors. In addition, the model proposed in this research can also be applied to other MOX gas sensor of the same type.
