Application of Machine Learning Algorithm for Mental State Attention Classification Based on Electroencephalogram Signals

This is the authors' abstract. We don't add key points for this paper.

Technological developments provide a new work environment where the human role is reduced to that of a passive observer. Risk in the workplace might result from a person's incapacity to maintain attention and concentration while doing passive control activities. Passive brain-computer interface (BCI) can be used to monitor mental attention status in humans (focused, unfocused, and drowsy) using electroencephalogram (EEG) signals. This study proposes using BCI to detect the level of mental attention status based on EEG signals using machine learning. In designing the architecture of this study, the EEG signal data has been decomposed by the Discrete Wavelet Transform (DWT) decomposition process with 4 decomposition levels and using the Daubechies family (dB). The signal decomposition results are extracted using several statistical features, which are then used for machine learning model features. The Xtreme Gradient Boost (XGBoost) algorithm was used to perform the classification task. The XGBoost model produced accuracy results of 99% (best) on individual subject tests and 98% (average) on all subjects.

Supports ML development for EEG attention classification; hypothesis-generating and requires prospective validation before clinical use.