Cloud-native microservice architectures are increasingly adopted in modern networked systems, including virtualized infrastructures and critical domains such as Healthcare 4.0. Ensuring their reliability requires timely anomaly detection and accurate root-cause analysis (RCA), despite challenges from service-level dependencies and indirect fault propagation. Traditional methods based on statistical thresholds or rule-based heuristics often lack the expressiveness to capture complex metricinteractions and treat RCA as a separate, post hoc task. We propose XTreeAD, an explainable, supervised framework for anomaly detection and root-cause localization in microservicebased systems. XTreeAD uses an XGBoost classifier to detect anomalous system states from monitored performance and resource utilization metrics, and applies SHapley Additive ex-Planations (SHAP) to compute per-service feature attributions. These are aggregated and combined with the service-call graph in a dependency-aware ranking algorithm to identify likely faulty services. Experiments on a public available dataset show that XTreeAD outperforms established baselines in detection precision, localization accuracy, and runtime efficiency.