Fatigue and situational awareness are key risk factors for pilot effectiveness in automatic cruise. In order to explore the nonlinear coupling relationship between pilot fatigue and situational awareness in automatic cruise and to reduce the multiple covariance between the predictive features, a comprehensive analysis based on machine learning interpretable techniques is performed in the spatial and temporal dimensions to improve the prediction accuracy. First, multidimensional time series were constructed from multimodal physiological data, and high-precision prediction models for both were built separately. The Shapley additive explanations were utilized in the spatial dimension to visualize the contribution of predictive features, identify the key features, and explain the unique attributes of physiological activities at different levels of efficacy. Second, we reconstruct the predictive model and analyze the evolution logic of fatigue and situational awareness in the time dimension and elucidate their coupling through correlation analysis. Finally, we fused the key features of fatigue and situational awareness, dynamically weighted the performance states, and constructed a comprehensive prediction model of pilot performance under auto-cruise. A total of 40 subjects’ physiological data of 90 min each were collected for analysis, and the prediction model was constructed with XGboost to demonstrate the feasibility of the proposed method. In contrast, the proposed method produces more accurate and interpretable prediction results, which can effectively contribute to the development of human-machine ergonomics in aviation.