Data Processing and Feature Engineering for Stock Price Trend Prediction

In recent years, there has been a significant advancement in AI technologies and their widespread application across various aspects of life. Analyzing stock prices using machine learning techniques and data mining has become a research topic that has captured the attention of the scientific community. Stock price prediction could greatly facilitate the investors’ appropriate decisions, improves profitability and hence decreases possible losses. This study aims at forecasting stock price movements using Vietnamese stock price data. We collected real-world price data of 18 stock codes from the Hanoi Stock Exchange platform. After checking for potential issues and having discovered critical data inaccuracies, we then applied various techniques to clean and augment the curated dataset. Employing technical analysis theory based on technical indicators, we proposed and implemented new features that improved prediction results by 9–20% over conventional feature generation methods. Our results are also of 2–11% better than those of some related studies on the same topic. Lastly, our data processing method achieved an accuracy of 75–80% in predicting results for the subsequent 6 months.