Electrotactile stimulation involves the direct stimulation of tactile receptors with a small electrical current and has advantages such as device miniaturization and high responsiveness. However, there are large individual differences in the sensations generated, which requires each user to adjust the current intensity before experiencing tactile sensation. This is a common disadvantage of electrotactile stimulation and hinders its practical use. In this study, we propose a method for measuring skin impedance (i.e., resistance (R) and capacitance (C)) in real time and estimating the individual stimulated current using machine learning. We measured skin impedances by fitting the voltage waveform between the anodic and cathodic electrodes to an exponential curve when stimulating a constant current pulse. We used 0.2 and 0.4 mA of prepulses (before stimulation) to estimate the electrical current of sensation threshold during cathodic stimulation. Results confirmed that machine learning can be used to estimate the stimulated current, and random forest regression was the most appropriate method (mean correlation coefficient of r2 = 0.95). The machine learning model was tested on 10 participants, which showed that the sensation threshold varied from 0.8 to 1.4 of the estimated value.