<p>In recent years, the integration of machine learning (ML) with the materials genome initiative has accelerated advancements in materials informatics, transforming the traditionally intricate processes in materials science. This review explores the detailed application of ML algorithms, with a focus on both supervised and unsupervised learning techniques, to develop predictive models for materials’ performance. This facilitates efficient materials screening and the emergence of innovative materials. Key applications, such as regression models for predicting alloy properties and the adoption of deep learning for high-throughput semiconductor screenings, are highlighted. The review addresses prevailing challenges in the domain, including data scarcity, high dimensionality, computational overheads, and model interpretability issues. A significant component of the discussion entails a critical analysis of ML algorithm efficiency in diverse materials science contexts. This is complemented by an examination of tailored feature selection methods, data preprocessing techniques specific to materials informatics, and the identification of suitable evaluation metrics. Emphasis is placed on optimizing the role of ML in materials science, especially its prowess in unveiling novel materials while navigating data limitations and model interpretability concerns. Looking forward, there is a strong inclination towards models that champion interpretability, incorporate domain-specific knowledge, and leverage innovative data generation strategies. In sum, this review serves as a comprehensive resource for researchers, elucidating key aspects, practical implementations, and potential pathways for ML applications in computational materials science.</p>