Methods like Electroencephalography (EEG) and magnetoencephalogram (MEG) record brain oscillations and provide an invaluable insight into healthy and pathological brain function. These signals are helpful to study and achieve an objective and early diagnosis of neural disorders as Developmental Dyslexia (DD). An atypical oscillatory sampling could cause the characteristic phonological difficulties of dyslexia at one or more temporal rates; in this sense, measuring the EEG signal can help to make an early diagnosis of DD. The LEEDUCA study conducted a series of EEG experiments on children listening to amplitude modulated (AM) noise with slow-rhythmic prosodic (0.5–1 Hz) to detect differences in perception of oscillatory ampling that could be associated with dyslexia. The evolution of each EEG channel has been studied in the frequency domain, obtaining the analytical phase using the Hilbert transform. Subsequently, the cause-effect relationships between channels in ach subject have been reflected thanks to Granger causality, obtaining matrices that reflect the interaction between the different parts of the brain. Hence, each subject was classified as belonging or not to the control group or the experimental group. For this purpose, two ensemble classification algorithms were compared, showing that both can reach acceptable classifying erformance in delta band with an accuracy up to 0.77, recall of 0.91 and AUC of 0.97 using Gradient Boosting classifier.