At present, resting state functional magnetic resonance imaging (rfMRI) has provided an efficient, rapid and advanced technology for brain function detection. Entropy can capture the dynamic characteristics of neural signals and might be used as a quantitative evaluation parameter. However, there are some problems remain solved yet, such as the entropy model computing with a fixed scale, and whether the entropy model could evaluate the cognitive performance. These problems will affect the accuracy of detection. Therefore, the multi-scale entropy model combined with a machine learning method is proposed here to investigate the relationship between complexity derived from