PREDICTING STOCK PRICE TRENDS USING MACHINE LEARNING METHODS WITH CONTINUOUS AND BINARY DATA

Abstract

Stock market trend has always been confusing and a challenging for investors and researchers because of several relevant factors. This study aims to robust machine learning model to predict stock price movement with acceptable degree of accuracy. From May 2016 to May 2021, daily historical records of four companies were collected for five years. We chose four of these firms at random: Amazon Inc., Brookfield Asset Management Inc., JP Morgan Chase & Co., and Exxon Mobil Corporation. Each of these firms can represent a different stock market industry, such as technology, real estate, financial, or energy respectively. Furthermore, raw data was cleaned and turned into continuous and binary data. This paper used six machine learning models (CART, Naïve Bayes, KNN, SVC, Logistic Regression classification, Random Forest) as a predictor. Experiment demonstrates that Logistic Regression classification and Random Forest classification gave us with reliable predictions on the movement of stock price continuous and binary data, respectively. Furthermore, models of continuous data may evaluate more accurately than models with binary value.