<p>Previous studies have demonstrated the positive effects of repetitive transcranial magnetic stimulation (rTMS) on the memory function of patients with mild cognitive impairment (MCI). However, the efficacy of rTMS exhibits individual variability, and the reasons for which (and the potential neural mechanisms) remain unclear. The aim of this study was to investigate neuroimaging biomarkers that can be used to predict the efficacy of rTMS treatment in MCI. Resting-state functional magnetic resonance imaging data were collected from 53 participants with MCI before and after the completion of 10-day high-frequency rTMS. Baseline intra- and inter-functional connectivity of the resting-state networks of responders and nonresponders was computed and compared to identify potential neural predictors of the variability in response to rTMS. The predictive value of individual and combined neural predictors was also assessed. Baseline intra-functional connectivity in the right parahippocampal gyrus of the hippocampal network (HN) and left precuneus of the default mode network (DMN) was significantly different between responders and nonresponders, as was inter-functional connectivity between the HN and the salience network (SN). These intra- and inter-functional connectivities were significantly correlated with the changes in recognition memory after rTMS seen in patients with MCI (<i>r</i> = −0.564, −0.322, and 0.289). The combination of these indicators showed the best classification performance for responders and nonresponders (sensitivity = 94%, specificity = 86%). These results suggest that functional connectivity within the HN and DMN, and between the HN and SN, is predictive of the treatment response to rTMS.</p>