Enhancing market trend prediction using convolutional neural networks on Japanese candlestick patterns

Area of Science:

• Quantitative Finance
• Machine Learning in Finance
• Financial Market Analysis

Background:

• Japanese candlestick patterns offer insights into market sentiment and trend direction.
• Technical analysis in financial markets relies on identifying patterns for predictive insights.
• Predicting future price movements is crucial for trading strategies.

Purpose of the Study:

• To predict future financial market price movements using Japanese candlestick patterns.
• To develop and train a Convolutional Neural Network (CNN) model for enhanced prediction accuracy.
• To validate the predictive capabilities of the CNN model through rigorous testing.

Main Methods:

• A structured three-step data preparation process was employed, including sliding window techniques and pattern identification using the Ta-lib library.
• Technical indicators were used to validate trend direction for each data window.
• A CNN model was developed to extract features from sub-charts for precise prediction.

Main Results:

• The developed CNN model demonstrated a high predictive accuracy of up to 99.3% for candlestick trends.
• Cross-validation techniques were implemented to ensure model reliability and performance on unseen data.
• The study successfully predicted directional movements of subsequent financial candlesticks.

Conclusions:

• Convolutional Neural Networks are effective tools for predicting financial market trends based on Japanese candlestick patterns.
• The proposed methodology provides a robust framework for enhancing the accuracy of technical analysis in trading.
• Accurate prediction of market sentiment and trend direction can be achieved through advanced machine learning techniques.