<p>Machine Learning algorithms role in network threat detection are playing a pivotal role which can be improvised with opimized solution for generalization, stability and predictive performance across such complex datasets. In this study, we are proposing a Hybrid Metaheuristic-based optimization framework that integrates the Grey Wolf optimization (GWO) algorithm with Decision Tree (DT) classifier for a concrete feature-parameter tuning and ensemble construction. The framework employs a multi-level optimization approach that utilized GWO algorithm which optimizes hyper parameters for DT machine learning model and model configuration with cross validation feedback during the optimization process. we also ensure generalization with fairness of process with five-fold stratified cross validation process that executed during the optimization process. This Hybrid model leverages the strengths of DT with GWO efficiently with tuning and exploration of hyper-parameter search space. Experiment results shows that proposed GWO-driven optimization significantly improves classification in both case binary and multi classification with respective matrices accuracy, precision, recall, F1-score, and RoC-AUC with comparatively grid or random search approaches. This article mainly highlights the efficiency of meta-heuristic-enabled algorithm optimization as a alternative to conventional tuning approach for constructing highly optimized, adaptive learning system. We have applied GWO_DT for binary classification as well multi-classification and having impressive results.</p>