Stock Price Prediction

Summary

The article provides an overview of stock price prediction, discussing the challenges, methodologies, and theories behind predicting stock movements, and suggests that while perfect prediction is debated, there is room for predictability based on historical data and publicly available information.

Abstract

The increasing interest in high-return stocks like Tesla, Amazon, and Google has prompted a deeper look into stock price prediction. The article summarizes advanced machine learning models that have improved predictive accuracy due to increased computational power and data accumulation from cloud technologies. It outlines the steps in time series prediction, including data preparation, algorithm definition, training, and evaluation, as well as the importance of a sound trading strategy and risk assessment. The predictability of stock prices is contentious, with the random walk theory and efficient market hypothesis suggesting that future prices are unpredictable. However, empirical studies support the weak form of market efficiency, indicating that historical prices can inform predictions. The article also touches on behavioral finance, which posits that market agents are not fully rational, and cites anomalies that cannot be explained by fundamental reasons alone. The piece concludes by directing readers to further reading on methods and features useful in stock price prediction and a book recommendation for a deeper understanding of predictions in financial markets.

Opinions

The author acknowledges the complexity of financial time series due to their nonlinearity and dynamism, suggesting that advanced machine learning models are necessary for accurate predictions.
The efficient market hypothesis is questioned, with evidence supporting only the weak form, implying that stock prices may be predictable to some extent based on historical data.
The article implies that while no method has consistently beaten the market, there is potential for predictability, especially considering the weak form of market efficiency.
Behavioral finance is presented as a counterargument to the fully rational market agent theory, highlighting the influence of psychology and irrational factors on stock prices.
The author encourages further exploration into stock price prediction, indicating that there is valuable knowledge to be gained from existing literature and research.

Stock Price Prediction — A review for beginners

There is an increasing interest in stocks like Tesla, Amazon, and Google due to the high returns they provided some time ago. This made me curious about this topic and I made research about what has been done in this field. I summarized the works that I am intrigued by. I hope you enjoy it.

Predicting stock price movements has always been a challenging and exciting subject for investors. Financial time series are large in size, nonlinear and dynamic; hence it is a highly demanding task to understand their nature. However, with the advance in computer technologies and computational power, the complexity of the machine learning models is increased, resulting in a higher predictive accuracy thanks to the possibility of accumulation of large amounts of data with the adoption of cloud technologies.

The solution of a time series prediction problem follows certain steps; data preparation, algorithm definition, training, and forecasting evaluation. Additionally, two other steps may be helpful: trading strategy and money evaluation. The trading strategy involves the determination of the rules to enter and exit the market, the proportion of investment considering available capital, and the risk involved with a good fit to the real world.

A significant issue of a financial time series forecasting problem is the evaluation step since any misleading outcome can land the investors with financial losses.

Predictability

According to the random walk theory, future prices cannot be predicted regardless of the methodology, whereas the efficient market hypothesis says prices fully reflect all the information on the market. Malkiel (2003) argues that any daily price change is random, the previous day’s price change is immaterial and any new information is unpredictable, and so are the prices. Malkiel explains the efficient market hypothesis with these words: “equity prices adjust to new information without delay and, as a result, no arbitrage opportunities exist that would allow investors to achieve above-average returns without accepting above-average risk”. Fama claims that the price of a security is a good estimate of its intrinsic value at any time in an efficient market. He used statistical tools to test if there are serial correlations or consecutive price changes in the same direction. He concluded that selected securities cannot do better than randomly selected ones.

In a later study, Fama conducted empirical tests dividing the efficient market hypothesis into three categories: weak, semi-strong, and strong. The strong form indicates the situation that the market prices are reflected by the relevant information to which a certain investor group has access; the semi-strong form indicates the situation that the market prices are reflected by the publicly available information and the weak form indicates the situation that the market prices are reflected by historical prices only. Surprisingly, study results support the weak form, but are insufficient to do so for the other two forms. This conclusion paves a path to the prediction of stock prices in the sense that the stock prices are predictable if the efficient market hypothesis is in weak form. On the other hand, Lawrence claims in his study from 1997 that none of the previous methods can beat the market steadily so far.

Behavioral finance argues that the agents operating in a market are not fully rational. Shiller gives examples of the anomalies January effect or the day-of-the-week effect in his paper and claims that there is no fundamental reason behind these occurrences. Still, they occur because of ‘sunspots’ or ‘animal spirits’ or just mass psychology.

This was the first post of my Stock Price Prediction series. You can read my second and third one below.

If you want to learn more about stock price prediction, check this book!

Thank you!

This post may contain affilliate links.

Stock Price Prediction — A review for beginners

Predictability