Machine learning prediction of electron density and temperature from He I line ratios

We propose to utilize machine learning to predict the electron density, n_e, and temperature, T_e, from He I line intensity ratios. In this approach, training data consist of measured He I line ratios as input and n_e and T_e measured using other diagnostic(s) as desired output, which is a Langmuir probe in our study. Support vector machine regression analysis is, then, performed with the training data to develop a predictive model for n_e and T_e, separately. It is confirmed that n_e and T_e predicted using the developed models agree well with those from the Langmuir probe in the ranges of 0.28 × 10¹⁸ ≤ n_e (m^-3) ≤ 3.8 × 10¹⁸ and 3.2 ≤ T_e (eV) ≤ 7.5. The developed models are, further, examined with an evaluation data, which are not included in the training data, and are found to well reproduce absolute values and radial profiles of probe-measured n_e and T_e.