The agricultural sector remains essential to preserve worldwide food security together with economic stability especially in territories where farmers heavily depend on agriculture for financial stability. The necessity to predict crop yields accurately increases because of population growth combined with changing environmental conditions. The research paper provides an organized review of crop yield forecasting systems which implement machine learning algorithms. The research evaluates supervised and ensemble learning models such as Random Forest (RF), Support Vector Machine (SVM) and Gradient Boosting (GB) and Artificial Neural Networks(ANNs) as well as advanced Deep Learning (DL) approaches to solve agricultural system complexities. The analysis shows that innovative ML algorithms need development for diverse high-dimensional data sets which include weather patterns and soil characteristics and crop types and historical yield records. This research analyzes recent advancements of remote sensing systems and IoT-based data collection regarding their functions for live accurate prediction operations. This research highlights the necessity of feature engineering together with data preprocessing alongside hybrid modeling strategies for solving issues that affect datasets and environmental factors as well as scalability. Different algorithms perform effectively for yield prediction purposes because their analysis has demonstrated high accuracy and robust prediction capabilities. The research provides detailed information to assist both practitioners and researchers who want to improve decision systems in agriculture with sustainable machine learning approaches for crop management.