Intrusion detection system is an essential component of network security protection mechanisms. Most intrusion detection system can not adapt the variation of network environment. Aiming at this problem, an adaptive strategy that composed of condition space and strategy space is proposed. The condition space describes the network environment and the strategy space describes the strategy. There is an exclusive strategy corresponds to a certain environment state in condition space. On the base of the adaptive strategy, an adaptive intrusion detection system based on agent and data mining is designed. The simulation experiments indicate that the adaptive strategy is effective.