The field of eXplainable Artificial Intelligence (XAI) has evolved to bridge the gap between complex, obscure black-box algorithms and human users. These methods try to find what matters most to the models and how it matters. Feature importance plays a pivotal role in answering this question, signifying the degree to which each feature contributes to the model's final output. However, existing approaches for estimating feature importance face limitations by often overlooking intricate relationships between input features. Challenges arise, particularly when input features are correlated, leading to potentially misleading results. To address these challenges, the causal knowledge among input features is leveraged to estimate feature attribution. However, establishing the knowledge of causal graphs may be difficult in real-world applications. Additionally, none of the current methods address the utilization of knowledge about the relationship among input features to handle correlation in cases where causality may not be present. Recognizing this gap, where methods either neglect or explicitly consider causal knowledge among input features, we propose a novel approach: MYerson additive Explanations on gRaphs (MYER). MYER aims to explain machine learning (ML) models by integrating network knowledge among input features without explicitly relying on causal assumptions. Leveraging the Myerson value and the network among input features, MYER calculates feature importance, presenting a framework that builds upon the groundwork laid by SHapley Additive exPlanations (SHAP) in estimating feature importance. The practical application of our proposed approach is demonstrated using feature importance as the indicator of health monitoring and fault diagnostics for gas turbine engines. This graphical approach allows users to comprehend the effect of abnormal behavior in a sub-component over the downstream sub-components.