Regarding gas sensors, in addition to diversifying detection methods, research is being conducted to improve the detection accuracy through machine learning. However, if the data quality is poor, machine learning does not lead to improved detection accuracy. However, one factor that can have an unexpected effect on gas sensor data is the presence of moisture in air. Gas molecules have an indicator of their affinity for water, called polarity. If the target gas is polar, it may be affected by water before it reaches the gas sensor, thereby causing the sensor output to change. However, if the target gas is not polar, it is less likely to be affected by water. We developed a grid that generates an electric field near the gas sensor. First, we verified whether we could obtain data that could be used for decision making using only the grid and sensor. We then improved the quality of the data so that they could be used for machine learning. In this study, we verified whether the gas sensor output differs between polar and nonpolar gas molecules depending on the strength of the electric field formed by the grid in the gas flow path. As a result, the gas sensor output for polar gases had a rounded waveform regardless of the effect of the grid. However, the gas sensor output waveform for nonpolar molecules was rectangular, regardless of the effect of the grid. In addition, the gas sensor outputs for the nonpolar and weakly polar molecules decreased while maintaining their respective waveforms. The results of this experiment suggest that the combination of a grid and a gas sensor may enable the identification of gases based on their polarity.