The invention relates to algorithm processing, and provides a Python program based implementation method for dynamic import of user-defined algorithm aiming at the problem that the requirement of user-defined algorithm processing data of a client cannot be met by a cloud system in the conventional technology, so that a user can dynamically add a self-defined algorithm. The method comprises the following steps: a. defining new algorithm storage data type, and changing actual algorithm procedure contents into an algorithm path; b. publishing algorithm file input and output requirements, whereinby regarding an algorithm procedure as a black box, a system only provides input and receives output; c. opening a Django file transmission interface to be used for receiving user-defined algorithm file and storing; d. recording an algorithm storage path to a storage data type; and e. when an algorithm needs to be called, dynamically import the algorithm procedure, processing data input algorithm,and outputting a processing result. The method is applicable to algorithm self-definition and dynamic calling.