DPP: Deep predictor for price movement from candlestick charts.

Area of Science:

• * Quantitative Finance
• * Machine Learning
• * Financial Time Series Analysis

Background:

• * Stock market price forecasting is complex due to numerous influencing factors like economic conditions, political events, and market news.
• * Traditional forecasting methods often struggle to capture the intricate patterns within historical stock data.
• * Candlestick charts offer a rich visual representation of price movements, yet their full potential in predictive modeling remains underexplored.

Purpose of the Study:

• * To propose a novel framework, the Deep Predictor for Price Movement (DPP), for enhanced stock market price forecasting.
• * To leverage candlestick chart patterns through a deep learning approach for improved predictive accuracy.
• * To evaluate the performance of the DPP framework against established forecasting models.

Main Methods:

• * Decomposing candlestick charts into smaller sub-charts for detailed analysis.
• * Utilizing a Convolutional Neural Network (CNN) autoencoder to extract optimal representations of these sub-charts.
• * Employing a Recurrent Neural Network (RNN) to predict future price movements based on the learned sub-chart representations.

Main Results:

• * The DPP framework was evaluated using trading data from the Taiwan Stock Exchange Capitalization Weighted Stock Index and the Nikkei 225 index.
• * Performance was benchmarked against established models including IEM, Prophet, and Long Short-Term Memory (LSTM).
• * Preliminary results indicate the DPP framework's potential for accurate stock price movement prediction.

Conclusions:

• * The proposed Deep Predictor for Price Movement (DPP) framework offers a promising new approach to stock market forecasting.
• * Integrating CNN-autoencoders and RNNs with candlestick chart analysis provides a powerful tool for capturing complex market dynamics.
• * Further research and validation on diverse datasets are warranted to solidify the DPP's efficacy in real-world financial applications.