From Technical Indicators to Machine Learning: A Data-Driven Approach to Price Movement Forecasting (Part 1.)
In this article I will try to replicate a proposed ML framework used for stock price movement predictions (Lin et al 2021). This framework uses Technical Analysis, custom K-Line patterns and historical price data for predictive modeling.
The objective of this project is to replicate several machine learning-based algorithmic trading strategies described in peer-reviewed publications. Specifically, I have focused on analyzing and reconstructing the approach proposed by Lin et al. (2021) for stock trend prediction using candlestick charting. To achieve this, I have utilized similar technical indicators and standard machine learning models, such as Random Forest, Support Vector Machines, K-Nearest Neighbors, Gradient Boost, and XGBoost, to forecast stock price movements within a 3–5 day window. I have also assessed model performance by the following metrics: Accuracy, Precision, Recall, F1 Score.
Through my work, I have discovered that the models in this project outperform the results presented by Lin et al. (2021) in multiple aspects. However real-world applicability is yet to be assessed (see Results).
0 Comments