Presentation Information
[10a-N323-9]Real-time Analysis of Neutral species spectrum Using Machine Learning (2)
〇(M2)Kotaro Takahashi1, Yusuke Ando1, Takayoshi Tsutsumi2, Kenichi Inoue2, Makoto Sekine2, Kenji Ishikawa2 (1.Nagoya University, 2.Nagoya Univ. Center for Low-temperature Plasma Sciences)
Keywords:
plasma process,machine learning
We developed a non-invasive and real-time analysis method for measuring neutral radical species in plasma, using optical emission spectra and machine learning. In our previous work, we demonstrated that random forest regression enables high-accuracy prediction of CH3 radicals (m/z = 15).In this study, we further introduced Partial Least Squares (PLS) regression to select and reduce the dimensionality of explanatory variables. The reduced features were then used as input for the random forest regression model, resulting in improved prediction accuracy compared to the conventional approach. This demonstrates that our method enables highly accurate and fast prediction of radical densities, and it can serve as an effective tool for the optimization of plasma processes.