The invention discloses an enterprise public information model extending method and system based on machine learning, and the method comprises the steps: building an enterprise public information physical model database, calculating the number of existing classes, and storing the data information of the existing classes; determining the attributes in an extension class; randomly extracting K existing classes as initial neighbor classes, and building a priority queue comprising the initial neighbor classes and the similarity between the initial neighbor classes and the extended class; carryingout the traversal calculation of the similarity between the existing classes, except the initial neighbor classes, and the extending class, and finding out K existing classes with the largest similarity from the existing classes to form a new priority sequence; carrying out the calculation of classes in a packet where the classes in the new priority sequence belongs, selecting the packet with thelargest number of classes as the packet where the extending class belongs, and writing the extending class into the packet. The method and system can achieve the quick, convenient and precise searching of the packet where the extending class belongs.