Obssesive-compulsive disorder (OCD) is a serious mental illness that affects the overall quality of the patients' daily lives. Accurate diagnosis of this disorder is a primary step towards effective treatment. Diagnosing OCD is a lengthy procedure that involves interviews, symptom rating scales and behavioral observation as well as the experience of a clinician. Discovering signal processing and network based biomarkers from functional magnetic resonance scans of patients may greatly assist the clinicians in their diagnostic assessments. In this paper, we explore the use of Pearson's correlation scores and network based features to predict if a subject has OCD. We extracted mean time series from 112 brain regions and decomposed them to 5-frequency bands. The mean time courses were used to calculate the Pearson's correlation matrix and network based features for each band. Minimum redundancy maximum relevance feature selection method is applied to the Pearson's correlation matrix and network based features from each frequency band to select the best features for the final predictor. A leave-one-out cross validation method is used for the final predictor performance. Our proposed methodology achieves 80% accuracy (23 out of 29 subjects classified correctly) with 81% sensitivity(13 out of 16 OCD subjects identified correctly) and 77% specificity (10 out of 13 controls identified correctly) using leave-one-out with in-fold feature ranking and selection. The most discriminating feature bands are 0.06-0.11 Hz for Pearson's correlation and 0.03-0.06 Hz for network based features. The high classification accuracy indicates the predictive power of the network features as well as carefully chosen Pearson's correlation values.