The invention provides a feature tensor-based Chinese knowledge graph representation learning method. The method comprises the following steps: carrying out data preparation, establishing a data structure, constructing an entity feature vector matrix, definiing relation vectors and distance formulas of marked triples, obtaining a training set, training a knowledge graph representation learning model, updating model parameters and iterative training, and carrying out relationship prediction on the unmarked triad by using the model, and carrying out iterative training again until a new unmarkedtriad cannot be learned. According to the method, feature tensors are formed by using Chinese pinyin, character information, word information and description information and are converted into featurevectors, so that a method for randomly initializing entity vectors in traditional knowledge representation learning is replaced, and Chinese features are fully utilized. Besides, a double-layer iteration mode is adopted to supplement the training corpus, so that the relation matrix can be continuously corrected, and the precision and convergence speed of the knowledge graph representation learning model are improved.