The solid-state battery (SSB) is a promising direction to address the inherent safety problems in commercial batteries and energy storage systems. However, the development of SSBs is still hider of the low ionic conductivity of solid-state electrolytes. Based on a machine learning (ML) method, a cobalt-doping strategy was developed for the NaZrSiPO (NASICON) compound by training on NASICON-type solid electrolyte data. The cobalt-doping strategy efficiently improves the NASICONs’ ionic conductivity to ∼2.63 mS/cm with low activation energy at ∼0.245 eV. The grain-boundary ionic conductivity reaches ∼11.00 mS/cm without extra densification of the pellet. The NASICON's structures were studied by the Rietveld and the bond-valence methods. The calculations and observed structural transitions confirm that the cobalt-doping strategy promotes the structural transition and adjusts the structure to a better performance state. The doping strategy predicted by the ML model is consistent with our experimental results, providing very useful guidance for improving ionic conductivity of NASICON electrolytes.