This paper addresses the historical issues and current difficulties associated with forecasting tropical cyclones, covering axisymmetric structures, dynamic dynamics, and forecasting methods. To overcome predicting constraints, researchers are increasingly using machine learning, a type of artificial intelligence. Machine learning techniques have the potential to improve forecasts related to tropical cyclones, including origin, track, intensity, extreme weather, and storm surge. These techniques can be applied to numerical models or data-driven models. In terms of predictability, stability, and utilizing a variety of data sources, the complexity of tropical cyclones presents both opportunities and challenges, even though there is unrealized promise in using machine learning for more accurate predictions.