Shear strength of reinforced concrete (RC) columns is an important evaluation indicator that is frequently used to determine if a RC frame is able to avoid global collapse under expected seismic loads. Many researchers have proposed various formula-based models to evaluate the shear capacity of both RC rectangular- and circular-section columns. However, the verification of these models has only been performed on a relatively small dataset of RC specimens (i.e., dozens of samples). Furthermore, the data samples used to develop the models are also the only specimens utilized to test the performance of the proposed models, excluding specimens which are not included in the data set used for model formulation. As an alternative, machine learning-based approaches have recently been implemented within the realm of structural engineering and have been validated effective. In general, machine learning-based approaches separate the training set (data/specimens used to develop a model) from the testing set (data/specimens used to evaluate the developed model). In this way, machine learning-based approaches can provide a robust and increasingly accurate alternative to traditional formula-based modelling strategies. In this work, the application of machine learning-based models to predict the shear strength of RC columns is presented. Various available machine learning-based models are developed to predict the shear strength for RC circular-section columns, and their prediction performances are compared. A database including 161 RC circular section columns is established. The performance of these machine learning-based models is examined through a 10-fold cross validation procedure, and further compared with results of popularly used formula-based models.