The invention relates to a processing method of air big data. The processing method comprises the steps of periodically traversing multiple content information sensed in a preset time period, affirming same information segments which continuously occur, processing the affirmed same information segments which continuously occur so as to reduce the size of the content information, and recognizing the content information, so as to construct initial information. By combining with the processing method, the defects in the prior art that the volume of a first message transferred from an air processing device to a server is large, the cost is high, the cost lowering difficulty is lowered, the subsequent index fitting accuracy is reduced, and the errors are very big after fitting when big time delay is large are effectively avoided.