As a method of modeling fusion plasma using a large-scale analysis database, we have achieved results on two new approaches: effective modeling by regression not based on plasma physics, and the introduction of data assimilation techniques into fusion research, which has resulted in the use of regression equations to express the gap between a transport model based on plasma physics and a model assimilated with observational data. In this regard, we aimed to obtain a predictable regression equation and to extract important variables based on a statistical perspective different from that of plasma physics, and used the Akaike Information Criterion (AIC) in the analysis using JMP. We were able to derive interesting results, such as a variable that was recognized in the experiment but not explicitly included in the model being selected as a statistically important variable, and the exponent of the variable was not inconsistent with that recognized in the experiment, which also leads to the complementarity of statistically important variables and plasma physics. We would like to report on these results.
