This paper proposes a learning-based anomaly classication of the guided-missile for performance evaluation. Identifying the anomaly of a missile from the performance metrics is difficult while analyzing the
detailed parameters from the Monte Carlo simulation requires expertise in the eld. Inspired by recent studies, machine learning is applied to extract features from time-series trajectory data to evaluate the performance of the guided missile. 6-DOF simulation data are used to train the model, and four network structures are tested for the anomaly classication problem. The results show that utilizing the time-series feature from the data enhances the classication accuracy and relational learning reduces the network size signicantly.