IQ level prediction and cross-relational analysis with perceptual ability using EEG-based SVM classification model

Abstract

This paper presents IQ level prediction and cross-relational analysis with perceptual ability using EEG-based SVM classification model. The study hypothesized that measure of perceptual ability and intelligence is strongly connected through the brain’s attention regulatory mechanism. Therefore, an intelligent classification model should be able to predict and map IQ levels from a dataset associated with varying levels of perception. 115 samples of resting EEG is acquired from the left prefrontal cortex. Sixty-five is used for perceptual ability analysis via CTMT, while another fifty is used in the development of IQ level classification model using SVM. The mean pattern of theta, alpha and beta bands show positive correlation between perceptual ability and IQ level datasets. Meanwhile, the developed SVM model outperforms the previous ANN method; yielding 100% accuracy for training and testing. Subsequently, the classification model successfully predicts and mapped samples from the perceptual ability dataset to its corresponding IQ levels with 98.5% accuracy. Therefore, validity of the study is confirmed through positive correlation demonstrated by both traits of cognition using the pattern of mean power ratio features, and successful prediction of IQ level for perceptual ability dataset via SVM classification model.