In the light of realizing the self-learning operation mode recognition of air borne fire control radar, an optimized LSSVM algorithm based on grid search and K-fold verification is proposed. First, this paper extracts feature parameters from non-cooperative radar signals and establishes library base of radar signal as training sample in LSSVM model. Next, the paper applies grid search method in parameter optimization to realize model adjustment under the circumstance of uncertain sample range. In the sequel, the paper utilizes K-fold cross validation for realizing performance evaluation and reducing model error caused by sample randomness to improve generalization ability. The simulation results show that the recognition accuracy of VS/RWS/TWS/STT modes reaches 97%, thus having a good recognition performance and practical value of proposed method.