Bruxism is known as the rhythmical clenching of the lower jaw (mandibular) by involuntary contraction of the masticatory muscles (masseter muscles) together with parafunctional grinding of the teeth that usually occur during sleep. It affects patients’ quality of life adversely due to tooth wear, tooth loss, and pain and fatigue in the jaw muscles. It is a common condition that is difficult to diagnose and treat. Bruxism diagnosis is often made by monitoring electromyography (EMG) activity of the masseter muscles during sleep. In this study, fatigue and pain in lower jaw muscles are examined together with teeth grinding and clenching activities. 13 time- and frequency-related features that are widely used in the literature were extracted from EMG signals. Correlations between the features were determined through five different algorithms in addition to Artificial Neural Network, and the features with the highest correlation were grouped in different combinations. As a result of tests performed on these groups of features, most effective features to be used in surface electromyography (sEMG) signal analysis were identified. K-nearest Neighbor, Support Vector Machine and Artificial Neural Network algorithms are shown to increase the accuracy of bruxism diagnosis.