An Accurate Data Cleaning Procedure for Electron Cyclotron Emission Imaging on EAST Tokamak Based on Methodology of Machine Learning

TL;DR

This paper presents a machine learning-based data cleaning method for electron cyclotron emission imaging on EAST tokamak, significantly improving signal classification accuracy and reducing manual effort in data analysis.

Contribution

It introduces a novel, accurate data cleaning procedure utilizing SVM and decision trees tailored for ECEI data on EAST tokamak, enhancing preprocessing reliability.

Findings

Recognition rates of saturated signals: 99.4%

Recognition rates of zero signals: 99.86%

Recognition rates of weak signals: 99.9%

Abstract

A new data cleaning procedure for electron cyclotron emission imaging (ECEI) of EAST tokamak is constructed. Machine learning techniques, including SVM and Decision tree, are applied to identifying saturated, zero, and weak signals of ECEI raw data, which not only reduces the effort of researchers for data analysis, but also improves the accuracy of data preprocessing. To enhance the reliability of the procedure, proper training sets are sampled based on massive raw data from the experiments of ECEI on EAST tokamak. Window size of temporal signal, kernel function, and other model parameters are obtained after model training. Consequently, the recognition rates of saturated, zero, and weak signals in raw data are 99.4%, 99.86%, and 99.9%, respectively, which proves the accuracy of this procedure.