Prediction of Amazon’s Stock Price Based on ARIMA, XGBoost, and LSTM Models

Abstract

Finding the best model to predict the trend of stock prices is an issue that has always garnered attention, and it is also closely related to investors’ investment dynamics. Even the commonly used autoregressive integrated moving average (ARIMA), extreme gradient boosting (XGBoost), and long short-term memory (LSTM) have their own advantages and disadvantages. We use mean squared error (MSE) to judge the most suitable model for predicting Amazon’s stock price from many aspects and find that LSTM is the model with the best fitting effect and the closest to the real curve. However, the LSTM model still needs to improve in terms of performance so as to reduce the bias. We anticipate the discovery of more models that are apt for predicting stocks in the future.