Tren dan Perkembagan Supervised versus Unsupervised Learning

Abstract

The rapid development of artificial intelligence (AI) in the last decade has driven the widespread adoption of machine learning in various sectors, from healthcare and finance to manufacturing and education. Two main approaches dominate: supervised and unsupervised learning. Supervised learning leverages labeled data to build predictive models with high accuracy and strong generalization capabilities, but its application is often hampered by the need for expensive, time-consuming, and error-prone labeling processes. In contrast, unsupervised learning operates on unlabeled data to identify latent patterns, clustering, and novel data representations. This approach excels in large-data scale and exploration efficiency, but faces challenges in interpreting results, validating performance, and implementing them in real-world systems. This article presents a recent literature review (2020–2024) based on searches in Scopus, IEEE Xplore, and Google Scholar using the keywords “supervised learning,” “unsupervised learning,” “machine learning trends,” and “hybrid approaches.” The results show a significant trend toward integrating hybrid approaches such as semi-supervised learning, self-supervised learning, and a combination of transfer learning approaches, aiming to overcome the limitations of each paradigm. Furthermore, research gaps were identified regarding the need for robust evaluation methods for unsupervised learning, the development of adaptive models for dynamic data, and increased transparency in the learning process. Critically, this article highlights the importance of an interdisciplinary approach in designing hybrid algorithms that are not only efficient and accurate, but also interpretive and reliable for data-driven decision-making. Recommendations for future research directions focus on the creation of a standardized evaluation framework, the integration of AI with edge computing technology, and the exploration of the potential of hybrid learning in the context of real-time data.