Mobile laser scanning (MLS) is a favored information source for urban areas because of its capacity to gather quick and dense three-dimensional information, and classification studies of these large point clouds are carried out by enriching the data set collected. This study discusses a method for classifying an MLS point cloud by constructing a local neighborhood relationship with cylindrical neighborhood information and augmenting the dataset with geometric features in the absence of RGB images. Two benchmark datasets, TUM-MLS1 and Toronto-3D, were employed. The Random Forest (RF) classifier, which has been preferred in many researches for MLS classification, was built and assessed on eight classes of benchmark datasets for point-based supervised classification. As a result, we achieved 94.5% overall accuracy with only four geometric features for both datasets. When comparing our findings for dataset of TUM-MLS1 to those of a previous study, we found a 2.4% increase in overall accuracy.