Evaluation of two models for predicting Amazon stock based on machine learning

Abstract

With the fast growth of artificial intelligence and technology, the use of machine learning techniques in financial markets is gaining popularity. As a result, many opportunities arise, such as predicting future stock movements. Financial markets are complex and constantly evolving environments, so analyzing them can be challenging and interesting. There are no specific rules to predict or estimate the value of a stock in the stock market, so one can do stock price prediction by various methods. In this project, the stock price data of Amazon for the past five years, ‘Date’, ‘Starting price’, ‘Closing price’, ‘Highest price’, ‘Lowest price’, ‘Adjusted close price’, and 'Volume’ are all included. The data were obtained from Yahoo Finance to predict the future stock price. Two forecasting models, the Linear Regression model, and the Long short-term memory model were analyzed, and based on the comparison of Mean Absolute Error, it was concluded that the LSTM forecasting model was shown to be more effective for forecasting the time-series data category.