A Light Gradient Boosting Machine Based Effectiveness Prediction Method of Space Reconnaissance System

To address the problem of effectiveness evaluation of space reconnaissance systems with limited data sets, vast array of indexes and inadequate model interpretability, an interpretable effectiveness prediction method based on Light Gradient Boosting Machine(LightGBM) and the Shapley Additive exPlanations(SHAP) is proposed. The thorough evaluation process is given and the index system of space reconnaissance systems is built. Then LightGBM is used to establish a nonlinear mapping the model between the indexes and the effectiveness. Additionally, the model's hyperparameters are adjusted using Bayesian optimization to improve accuracy and lessen computational load. Furthermore, the feasibility of the proposed method is validated by modeling and evaluating the performance of the space reconnaissance system under a particular operational scenario. The experimental results show that compared with other mainstream machine learning methods, the proposed method has higher prediction accuracy and can efficiently extract the potential characteristics of the space reconnaissance system.