By 2020, 5G era will be commercially available. The smart city construction will also make great progress. Compared to current situation, more than thousand times of devices will connect to the cellular networks. For the operators, in order to analyze overall network performance, it is a key factor to estimate the user equipment (UE) radio frequency (RF) condition. However, practical RF estimation scheme is based on UE data log which can only observe UE that is at the top-serving cell with good RF condition. However, according to the comparison of actual UE data log and the scanner data log, potential RF problems may still exist since the UE will not always be served by the top-1 cell. In this paper, we propose a novel estimation scheme by integrating machine learning (ML) algorithm to analyze the scanner data logs from the target estimation zones where the mobility problems may occur. A hypothesis is obtained from learning step by various kinds of RF condition as input features. The numerical results show that the proposed estimation algorithm integrated ML can estimate probability of the potential mobility problems accurately.