With the complex wing ice accumulation problem caused by Supercooled Large Droplets (SLD), the impact of ice shape on flight safety has become increasingly significant. Machine learning technology exhibits considerable potential in the domain of numerical prediction for wing ice shapes. Machine learning-based prediction models can efficiently and accurately predict wing ice shape and reduce flight risk. This paper conducts a systematic re-view of the application of machine learning models in the realm of ice shape numerical prediction. It analyzes and contrasts the strengths and weaknesses of various machine learning models, such as Artificial Neural Networks (ANN), Support Vector Machines (SVM), and Gaussian Process Regression (GPR), alongside their adaptability in the ice shape prediction process. Additionally, this paper highlights the existing challenges and issues within these models, such as data quality and data sparsity, and provides a perspective on the research achievements, trends, and future directions in the field of ice accretion prediction using machine learning models.