Understanding and classifying phase transitionsPhase transitions in lattice-based systems remain central challenges in statistical physics and material science, particularly when conventional order parameters become ambiguous. Recent advances in machine learningMachine learning have shown promise in automating phase detection, yet many approaches either focus narrowly on identifying the critical temperature or rely on raw spin configurations, which can hinder interpretability and robustness. In this work, we develop a hybrid computational framework that integrates Monte Carlo simulations of the two-dimensional Ising modelIsing model with unsupervised machine learningMachine learning techniques to address these challenges. By extracting high-level thermodynamic observables—magnetization, energy, specific heat, and susceptibility—we construct a feature-rich dataset that captures both average behavior and fluctuation-driven response. Dimensionality reduction via Principal Component Analysis (PCA), followed by KMeans clustering, enables data-driven classification of ferromagnetic, paramagnetic, and antiferromagnetic regimes. The framework successfully identifies first-order transitions through hysteresis loops, resolves second-order critical phenomena near Tc \(\approx \) 2.269, and detects the Néel transition in antiferromagnetic systems where traditional magnetization fails. These findings highlight the novelty of using thermodynamic observables for interpretable ML-based phase classification and demonstrate the potential of this approach for analyzing smart materialsSmart materials and adaptive technologies.