Enhancing market trend prediction using convolutional neural networks on Japanese candlestick patterns
This study discusses using Japanese candlestick (JC) patterns to predict future price movements in financial markets. The history of candlestick trading dates back to the 17th century and involves the analysis of patterns formed during JC trading. Candlestick patterns are practical tools for the tec...
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
PeerJ Inc.
2025-02-01
|
| Series: | PeerJ Computer Science |
| Subjects: | |
| Online Access: | https://peerj.com/articles/cs-2719.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | This study discusses using Japanese candlestick (JC) patterns to predict future price movements in financial markets. The history of candlestick trading dates back to the 17th century and involves the analysis of patterns formed during JC trading. Candlestick patterns are practical tools for the technical analysis of traders in financial markets. They may serve as indicators of traders’ documents of a potential change in market sentiment and trend direction. This study aimed to predict the following candle-trend-based JC charts using convolutional neural networks (CNNs). In order to enhance the accuracy of predicting the directional movement of subsequent financial candlesticks, a rich dataset has been constructed by following a structured three-step process, and a CNN model has been trained. Initially, the dataset was analyzed, and sub-charts were generated using a sliding window technique. Subsequently, the Ta-lib library was used to identify whether predefined patterns were present within the windows. The third phase involved the classification of each window’s directional tendency, which was substantiated by employing various technical indicators to validate the direction of the trend. Following the data preparation and analysis phases, a CNN model was developed to extract features from sub-charts and facilitate precise predictions effectively. The experimental results of this approach demonstrated a remarkable predictive accuracy of up to 99.3%. Implementing cross-validation techniques is essential to verify the reliability and overall performance of the model. To achieve this goal, the dataset was divided into several small subsets. Subsequently, the model was trained and evaluated multiple times using different combinations of these subsets. This method allows for a more accurate assessment of the model’s predictive capabilities by examining its performance on unseen data. |
|---|---|
| ISSN: | 2376-5992 |