Candlestick Pattern Recognition Using Modified YOLOv8 in Financial Market Analysis

Pattern recognition in candlestick charts poses a formidable challenge due to the intricate shapes and intrinsic noise in financial data. This study addresses the critical need of a dataset to accurate pattern identification, with a specific focus on four highly used patterns: Head and Shoulder, Reverse Head and Shoulder, Double Top, and Double Bottom. The proposed modified YOLOv8 model excels in detecting patterns, achieving a mAP@50 value of 86.1%. The model's stability and consistency across epochs underscore its potential for real-time financial market analysis and short-term trading applications. This work pioneers the application of Object Detection models as an effective strategy for addressing pattern recognition challenges in financial markets, promising advancements in technical analysis.