Smart agriculture is a collection of techniques and technologies that aim to improve farming methods and production and support sustainable agricultural practices. This work proposes a Machine Learning-based Decision Support System (ML-DSS) for real-time decision support to farmers. The primary goal is to derive crop yield predictions, pest detections, and resource management through supervised machine learning models(es-implementation) using IoT-based sensor data. The architecture supports several machine learning techniques, including deep learning, ensemble models, and explainable AI frameworks, which can process heterogeneous data sources related to soil quality, weather conditions, and plant health indicators. A cloud-based platform is utilized for data collection, preprocessing, and predictive analytics. The experimental work is validated using real-world datasets from precision farming applications. Experimental results demonstrate significant overall prediction accuracy, improved decision-making speed, enhanced capacity for resource allocation, and reduced greenhouse gas emissions. Because of the use of interpretable AI techniques, model transparency has been facilitated, and trust from farmers is achieved. Finally, this research illustrates that the ML-DSS has the potential to increase agricultural productivity, moderate costs in the farmers' operations, and information-driven farming decisions for the future directions of adaptive learning.