Abstract
Stock price is an unstable time series affected by plenty of factors. Since various noises have significant impacts on its trend, the way to realize more accurate forecasts in terms of multidimensional data features has become a concern for scholars worldwide. Among all the methods, machine learning approaches are one of the popular ideas in recent years. This paper introduces the meaning of stock price prediction and the development of machine learning in this field for the past few years. Theoretical background of Random Forest, XGBoost and LSTM are provided and the state-of-art researches based on the above methods are also summarized. It concludes with a discussion of these models and the limitations of this paper, as well as an outlook for future work. The study aims to synthesize the scattered sources of information for the reference of later scholars. As a result, human beings can find better ways to maximize investment benefits and warn of stock market crises in years to come. Overall, these results shed light on guiding further exploration of stick price forecasting.