Predicting brain age from EEG signals using machine learning and neural network

Abstract

The objective of this study was to develop a technique for calculating the ages of people’s brains by analyzing EEG data signals and using machine learning algorithms on a Raspberry Pi. We employed many machine learning techniques, including random forest (RF), Decision Tree Classifier, K Nearest Neighbors Classifier (K-NN), Gaussian Naive Bayes, and Multi-layer Perceptron classifier(MLP). K-NN stands for K-nearest Neighbors, whereas RF stands for Random Forest. We initially implemented our machine learning algorithms on a desktop computer with many bells and whistles, where the dataset was also trained. By applying the Random Forest classifier (RF), we were able to attain 90% accuracy, the maximum feasible. The K-Nearest Neighbors classifier placed second with an accuracy of 87%. The accuracy obtained by the Decision Tree Classifier, the Naive Bayes algorithm, and the MLP algorithm, in order, was 83%, 39%, and 40%, respectively. Our major aim was the creation of an Internet of Things-based device, we tested our data on Raspberry Pi. If in the future, we were to construct, based on our model, a device that rapidly turned EEG brain signals into the participant’s brain age, we would be able to significantly improve the quality of our work. In addition, it will aid in the diagnosis of some brain illnesses at an early stage, which has been a struggle up until now.