Eyes-open (EO) and eyes-closed (EC) are the two experimental conditions during resting state functional magnetic resonance imaging (fMRI) scan sessions. However, the dynamic neural mechanisms of EO/EC based on intrinsic connectivity networks (ICNs) remains largely unexplored. This paper aimed to decode the dynamic internetwork neural mechanisms for EO/EC using data mining and to identify EO/EC resting state fMRI scans based on machine learning. To achieve these goals, the two states were analyzed using the discriminative models, resulting in total accuracy of 85.87%, a sensitivity of 91.3%, and a specificity of 80.43%. In addition, the discriminative features discovered using data mining were related to previous findings. In summary, we applied visual networkrelated inter-ICN features to decode the neural mechanisms of EO/EC. The reproducible results suggested that visual network-related inter-ICN dynamic features could be beneficial for decoding visual attentions, and had potential as neuroimaging-markers to identify EO/EC resting state fMRI scans. K E Y W O R D S data mining, dynamic functional connectivity, eyes-open/closed, neural mechanisms