EEG classification using Neural Network – An Application of Machine Learning in Classification of attention deficiency, to measure the effect of ChakaMarkaKosha Meditation-II

Stress reduces students’ attention span and is a common problem that contributes to students’ poor academic performance and self-efficacy. Various meditation techniques have been proven to help manage these challenges as they have improved college students’ focus, concentration, and performance. The author did a study on ChakraMarkaKosha Meditation (CM), a meditation on human energy centers that could reduce stress through Heart Coherence in his previous research. As a sequel, this study uses EEG technology to measure attention deficiency and proposes to improve it by the newly composed meditation (CM-II). The developed guided meditation focuses on reducing distractions and hyperactivity in individuals. The hypothesis is that CM-II improves attention, and increases students’ problem-solving skills and self-efficacy. Attention is one of the cognitive skills that involve concentration, problem-solving, judgment, and language. Attention-Deficit Hyperactivity Disorder (ADHD) is a neuro-developmental disorder that is characterized by hyperactivity and inattentiveness. Thus, an ADHD dataset is used in this study to detect a lack of attention using a Machine Learning classifying technique. Multi-layer Perceptron (MLP), a feedforward artificial neural network, is a fully connected multi-layer neural network that can be used in the classification of ADHD. Here, we used an existing EEG dataset that includes data collected from 30 kids who had been diagnosed with ADHD and 30 kids who were non-ADHD controls. To achieve classification for poor attention by electroencephalogram (EEG) as an objective approach, we used a feature extraction process using Fourier analysis and frequency band calculations. Alpha (α), Beta (β), Gamma (γ), Delta (δ), and Theta (θ) bands were extracted, and spectral analysis was done on theta/beta ratio (TBR) - an index of inattention. These techniques utilized for classification and spectral analysis were developed in Python, and after training and testing, it was found that the model produced 61.5% accuracy. This model will be used in the next phase to determine the effect of CM-II through live EEG data analysis in addition to self-efficacy analysis