Several studies have reported that Electroencephalograph (EEG) could provide a practical solution towards giving information with regards to brain functionalities. In the classification of EEG signals of dyslexic children, using a large number of electrodes would result in longer preparation time and noisy signals due to the uncomfortable feeling of wearing a cumbersome headset. This paper describes the localization of electrodes through classification of EEG signals into 3 groups; normal, poor dyslexic and capable dyslexic, using SVM with the aim to obtain the minimum number of electrodes without comprising its performance. Beta band power features were extracted using discrete wavelet transform (DWT) with Daubechies of order 2 and the classification algorithm uses a multiclass support vector machine (SVM) for the identification of subjects. Results showed that the classifier performance dropped 25% in the classification of capable dyslexic, 75% in poor dyslexic and 12% in normal when the electrode locations are localized to four (4) by removing C3, C4, T7 and T8. Performance accuracy also dropped when only (six) 6 locations were used by alternately removing first C3 and C4 and followed by T7 and T8. It can be said that the optimum electrode locations in the assessment of dyslexic would still have to be maintained to a minimum of eight (8).