In aquaculture process, common water quality monitoring system just collect data in real-time, and data-monitoring that have been collected is time postponed. It is necessary to conduct prediction and early warning on quality in accordance with its historical state and current state. This paper aims to conduct prediction and warning in terms of the DO content in carp aquaculture using neural network and decision tree, and try to complete the dissolved oxygen Real-time prediction and early warning system through prediction and early warning model research, system design, and system development. The effect in practical application shows that the system can use the two methods to predict DO content, and conduct early warning by value prediction and rule based reasoning.