Stock Market Prediction: Integrating Explainable AI with Conv2D Models for Candlestick Image Analysis

Joao Paulo Euko, Flávio Arthur O. Santos, Paulo Novais

Abstract: The term “candlestick” refers to a graphical representation of price movements in financial markets. Predicting candlesticks is crucial for anticipating market trends and making informed investment decisions. Interpretability is essential to understand the rationale behind model predictions, particularly in complex financial environments. In conclusion, this research leverages explainability techniques like SHAP and GRAD-CAM to improve the interpretability of Conv2D models for candlestick prediction in stock market images.