Schizophrenia is always a fascinating research area among the other psychological disorders due to its complexity of severe symptoms and neuropsychological changes in the brain. The diagnosis of schizophrenia mostly depends on identifying any of the symptoms, such as hallucinations, delusions and disorganized speech, entirely relying on observations. Researches are going on to identify the biomarkers in the brain affected by schizophrenia. Diverse machine learning approaches are applied to identify brain changes using fMRI studies. However, no conclusive clue has been derived yet. Recently, resting-state fMRI gains importance in identifying the brain's patterns of functional changes in patients having resting-state conditions. This paper aims to study the resting-state fMRI data of 72 schizophrenia patients and 72 healthy controls to identify the brain regions showing differences in functional activation using a two-stage feature selection approach. In the first stage, the study employs a novel mean-deviation-based statistical approach (Indranath Chatterjee, F1000Research, 7:1615 (v2), 2018) for voxel selection directly from the time-series 4-D fMRI data. This approach uses statistical measures such as mean and median for finding the significant functional changes in each voxel over time. The voxels showing the functional changes in each subject were selected. After that, considering a threshold 'α' on the mean-deviation values, the best set of voxels were treated as an input for the second stage of voxel selection using Pearson's correlation coefficient. The voxel set obtained after the first stage was further reduced to select the minimal set of voxels to identify the functional changes in small brain regions. Various state-of-the-art machine learning algorithms, such as Linear SVM and Extreme Learning Machine(ELM), were used to classify healthy and schizophrenia patients. Results show the accuracy of around 88% and 85% with SVM and ELM, respectively. Subtle functional changes are observed in brain regions, such as the parietal lobe, prefrontal cortex, posterior cingulate cortex, superior temporal gyrus, lingual gyrus, cuneus and thalamus. This study is the first-of-its-kind rs-fMRI study to employ the novel mean-deviation-based method to identify the potentially affected brain regions in schizophrenia, which eventually may help in better clinical intervention and cue for further investigation.