Abstract: fMRI data is typical small sample data with high-dimension,how to extract and select important features from high-dimensional data is the key for classification and recognition.Combined with the characteristics of fMRI data,a method of feature selection using Random Forest algorithm is proposed in this paper.To achieve the feature selection,Random Forest classification accuracy as a criterion to estimate the importance of features.The method was applied to healthy controls and patients with schizophrenia recognition,by calculating the contribution of each feature for classification to choose the features with greater contribution for classification and recognition,and locate to the corresponding brain regions according the order numbers of important features,providing an objective reference for clinical diagnosis.Experimental results show this method has a better performance.