The invention relates to an information classification processing system and method based on an urban IoT (Internet of Things), particularly relates to an information classification processing, prediction, decision and automatic learning method, and belongs to the technical field of IoT information processing. The system comprises a sensing layer, a data transmission layer, a data filtering layer,an information definition layer, an information processing layer, a decision supporting layer, a network layer and a decision tree autognosis method based on a self-built database; real-time sensed data is output at regular time by the data transmission layer; and the decision supporting layer carries out analysis statistics according to information uploaded by the information processing layer, combines instant information to carry out comparison with historical information, predicts future information which cannot be obtained in real time by a sensor and event trends, calculates an occurrence probability of an event along with a master variable at a future time point, makes a plurality of suggestion decision schemes, selects an optimal scheme to execute, collects process information andresult information and establishes a decision tree model, and after repeated optimization, the system automatically learns and carries out cognition, so that a decision is gradually made by the system.