This paper addresses the crew site visit scheduling problem whose purpose is to reduce the unplanned power outage causing by digging activities nearby underground cables. It is known that a supervisory crew visiting a digging event can significantly curtail the failure probability. However, due to limited workforce, a power distribution company cannot assign crew to visit all registered digging events. In this paper, a predictive model is deployed to obtain the failure probability for each event. A mathematical formulation and a branch-and-cut algorithm are introduced so as to give an optimal solution based on the predicted failure probability. A case study on a real-world electricity distribution network in the Netherlands is presented to show the efficiency of the proposed approach.