Predicting the fatigue life of materials accurately and efficiently remains a challenge due to complex loading conditions, multidimensional stress factors, and the nonlinear properties of composite materials. However, advancements in data-driven modeling, specifically machine learning (ML) techniques, have shown promise in mitigating these limitations. Different ML models have been extensively employed to forecast the fatigue life of materials. This review provides a comprehensive analysis of data sources, ML model types, and evaluation techniques utilized in the field of ML-based fatigue life prediction. It is important to note that all ML models comprise two fundamental elements: datasets and algorithms employed during the training process. The ultimate objective of any ML model is to achieve accurate predictions, which necessitates the optimization of hyperparameters. Hyperparameter tuning is performed using various techniques, as there is no universally recognized rule for maximizing ML model accuracy. The present state of research in this field emphasizes expanding fatigue datasets, refining hyperparameter tuning methods, and identifying influential factors to enhance prediction accuracy while optimizing computational time and cost. This paper aims to provide a comprehensive understanding of the advancements made in ML-based fatigue life prediction, shedding light on potential avenues for future research.