The traditional methodology employed within nowcasting systems involves an extrapolation scheme that either tracks storms as objects or looks for correlations between time periods to ascertain a storm motion vector. The novel method of using random forest classification for nowcasting takes a completely different approach. In machine learning, a random forest is a classifier that consists of many decision trees and outputs the class that is the mode of the classes voted by individual trees. In this paper, the random forest technique is used to nowcast oceanic convection using indicators derived from geostationary and polar-orbiting satellites and a global numerical weather prediction model. Mature oceanic convection is represented by a satellite-based indicator called the Convective Diagnosis Oceanic (CDO), which is treated as the truth field for random forest training/classification. The goal of nowcasting oceanic convection is thus converted to forecasting CDO intensity, which ranges in value between 0 and 4. As a first attempt of exploring the random forest technique in oceanic weather, seven days of data are used to train a forest of 200 decision trees, while three other days are used for independent classifications and verifications. An initial set of 16 predictors is used in the random forest training/classification process and includes various satellite and numerical model-derived fields. Both 1-hr and 2-hr CDO forecasts are created. The methodology of random forest classification, its advantages and disadvantages, and some nowcasting results obtained through random forest classification will be presented. Future work will also be discussed.