Daily Stock Price Prediction: A Case Study in Vietnam

Abstract

Stock price prediction remains a topic of debate. The Efficient Market Hypothesis argues against the need for prediction, stating that markets already incorporate all relevant information. However, the emergence of Deep Learning in Machine Learning has led many to believe in the potential of sophisticated algorithms in uncertain and volatile stock markets. The outcome of this debate varies based on market characteristics. To analyse this discussion within the framework of emerging market like Vietnam, a study examines nine stocks listed on the Ho Chi Minh Stock Exchange (HOSE). Time-series data is visualized and analysed to test various hypotheses. Three algorithms Linear Regression, SARIMA, and LSTM are researched and applied to predict stock prices. The predictions are compared and benchmarked against the Naive approach. The study concludes that the Linear approach outperforms Deep Learning. This difference may be attributed to the characteristics of the emerging market, such as its young age, low regulation, and high volatility with outliers. Traditional Machine Learning and Deep Learning both surpass the Naive approach, rejecting the Efficient Market Hypothesis in Vietnam.