As the key basis for evaluating the success or failure of air defense operations, it has become an urgent task to evaluate the operational effectiveness of systematic air defense weapons. To address the issues of high dimensionality, strong complexity, and subjective evaluation methods in the combat effectiveness indicators of air defense systems, a combat effectiveness indicator system for air defense systems based on the effectiveness environment is constructed through in-depth analysis of the observing, locating, deciding, and acting (OODA)environment combat theory. By fully utilizing the massive data generated during air defense operations, and extracting information features from them, a combat effectiveness evaluation method based on machine learning deep belief network (DBN) is proposed, and multiple methods such as Back Propagation ༈BP༉neural network, Particle Swarm Optimization - Back Propagation ༈PSO-BP༉neural network, Support Vector Regression ༈SVR༉algorithm, Deep neural networks ༈DNN༉network are applied for comparative simulation verification. The simulation results show that the predicted evaluation results of the DBN network model are closer to the actual values, fully proving the effectiveness of the machine learning model in the systematic evaluation of air defense combat effectiveness, achieving accurate evaluation of air defense system combat effectiveness, and providing strong support for commander combat decision-making.