The invention relates to a semi-supervised classification method based on probabilistic neighbors. The semi-supervised classification method comprises the steps of preparing a data set, preprocessingthe data set, constructing a probabilistic neighbors matrix S, initializing a probabilistic transition matrix P into a probabilistic neighbors matrix S, and performing label propagation, updating, inspection convergence, classification result obtaining and the like on an original data set. According to the method, the problems that in existing graph-based semi-supervised classification, construction of similarity graphs cannot well reflect the actual situation and classification is inaccurate are solved, and the classification method can be closer to the actual situation and enables the classification result to be more accurate.