Intelligent Models for Stock Price Prediction: A Comprehensive Review

Introduction

The stock market is a platform or a mutual organization that provides a trader to buy or sell stock shares. They form one of the critical parts of a country's economy as it is an essential way for companies to raise capital (Billah, Waheed, & Hanifa, 2017). Businesses and corporations allowed to offer shares to the public are termed public listed companies, and they have a significant impact on the economies in which they operate (Pun & Shahi, 2018). In most modern economies, other business organizations heavily rely on the funds generated by these financial markets. Therefore, analyzing the behaviour and performance of these financial markets has become a crucial research field. These analyses may include but are not limited to predicting prices of securities such as stocks, bonds, foreign exchange rates, market indicators, and trading volumes (Samarawickrama & Fernando, 2018). The stock market attracts investors and investment institutions' attention due to its high returns (Yao, Luo, & Peng, 2018). Most investors' goal is to predict the stock market's associated risk to decide between buying or selling shares of stocks while seeking to maximize profit on investment. However, predicting the behaviour of stocks is difficult because the market is highly volatile and influenced by unmeasurable external factors, including the global economy, events, politics, and investor expectations (Oncharoen & Vateekul, 2018). Stock markets are considered the heart of the world's economy, in which billions of dollars are traded every day. The correct prediction of the future behaviour of markets would be extremely valuable in various areas (Hoseinzade & Haratizadeh, 2019). Traditionally, several conventional methods based on time series have been proposed to aid in predicting the stock market. Classical models like the Black-Scholes has also been used to model the stock market in predicting its volatility. Despite the many works done, accuracy still remains a challenge in this domain. Presently, these markets are known to have generated enormous data, which is of interest to the data science community. With the deep drive of intelligence, machine learning has played a useful role in the prediction of stock price leading to the proposal of several efficient algorithms as discussed under the section of this study named “Papers Reviewed”. These learning algorithms learn from historical price data to predict future prices (Nelson, Pereira, & Oliveira, 2017). However, this historical data are expected to be clean as much as possible as a bit of tweak in the data can perpetuate massive differences in the outcome (Parmar et al., 2018). Taking into consideration the intricate nature of this domain and the diverse contribution made by the research community without it being properly synchronized, this study seeks to present a systematic review of what has happened in the past as a contribution to knowledge continuity. The study is organized as follow: we have the background that gives, in brief, an overview of machine learning, followed by research methods that explain the protocols adopted to carry out this study. Next, is papers reviewed where some selected papers are discussed. The discussed paper are analyzed in the result and discussion section, and finally is the conclusion and future works.