Abstract: AST3-2 (Antarctic Survey Telescopes) Telescope locates in Dome A, the loftiest ice dome on the Antarctic Plateau. It produces huge amount of observation data which requires more efficient data reduction program to be developed. Also data transmission in Antarctica is much difficult, thus it is necessary to perform data reduction to detect variable sources and transient sources remotely and automatically in Antarctica, but this attempt is restricted by the poor computer performance in Antarcitca.For the realization of this aim, developing a new method based on pre-existing image subtraction method and random forest algorithm, taking the AST3-2 2016 dataset as test sample becomes an alternative choice. This method performs image subtraction on data set, then applies principle component analysis to extract the features of residual images. Random forest is used as a machine learning classifier, and a recall rate of 97% is resulted. Our work verifies the feasibility and accuracy of our method, and finally finds out a batch of candidates for variable stars in the AST3-2 2016 dataset.