Stock Market Pattern Detection using YOLOv8

This project is designed to detect stock market patterns (e.g., candlestick patterns) in financial time-series data using YOLOv8, a state-of-the-art object detection model. It processes historical stock or forex data, generates candlestick charts, and detects patterns using a pre-trained YOLOv8 model. Detected patterns are saved for further analysis.

Features

• Candlestick Chart Generation: Converts time-series data into candlestick charts for visualization.
• Time Gap Filtering: Skips chart generation if there is a time gap greater than 10 minutes in the data.
• Pattern Detection: Uses YOLOv8 to detect stock market patterns in the generated charts.
• Automatic Saving: Saves charts and detected patterns into organized directories.
• Customizable Parameters: Allows users to define window sizes for chart generation and adjust detection confidence thresholds.

Requirements

You can install the required libraries using the following command:

pip install -r requirements.txt

Dataset

The project expects a CSV file containing historical stock or forex data with the following columns:

• time: Timestamp of the data.
• open: Opening price.
• high: Highest price.
• low: Lowest price.
• close: Closing price.

Example dataset format:

time	open	high	low	close
2023-10-01 12:00:00	1800	1810	1795	1805
2023-10-01 12:05:00	1805	1815	1800	1810

How It Works

1. Data Preprocessing:

   • The script reads the CSV file and processes the data into a pandas DataFrame.
   • It converts the time column to datetime format and sets it as the index.

2. Time Gap Filtering:

   • The script checks for time gaps greater than 10 minutes in the data.
   • If a time gap is found, the corresponding chart is skipped.

3. Candlestick Chart Generation:

   • The script generates candlestick charts for specified window sizes (e.g., 2 hours, 4.1 hours, 6 hours).
   • Charts are saved in the images directory.

4. Pattern Detection:

   • The script uses a pre-trained YOLOv8 model (foduucom/stockmarket-pattern-detection-yolov8) to detect patterns in the generated charts.
   • Detected patterns are annotated and saved in the detected_patterns directory.

5. Output:

   • Charts with detected patterns are displayed using OpenCV and saved.
   • Logs are printed to indicate whether patterns were detected in each chart.

Supported Patterns

• Head and shoulders bottom
• Head and shoulders top
• M_Head
• StockLine
• Triangle
• W_Bottom

Usage

Running Locally

1. Clone the repository:

   git clone https://github.com/abbasi0abolfazl/stock-market-pattern-detection.git
   cd stock-market-pattern-detection

2. Place your dataset (CSV file) in the project directory.

3. Run the script or pattern_detection.ipynb:

   python main.py

4. Check the output:

   • Generated candlestick charts are saved in the images directory.
   • Detected patterns are saved in the detected_patterns directory.

Running on Google Colab

You can also run this project directly on Google Colab:

1. Open the Colab notebook using the link above.
2. Upload your dataset (CSV file) to the Colab environment.
3. Run all cells in the notebook.
4. Check the output:
   • Generated candlestick charts are saved in the images directory.
   • Detected patterns are saved in the detected_patterns directory.

Customization

• Window Sizes: Modify the window_sizes list in the main function to change the time intervals for chart generation.
• Time Gap Threshold: Adjust the max_time_gap variable in the generate_candlestick_charts function to change the maximum allowed time gap.
• Model Parameters: Adjust the YOLOv8 model parameters (e.g., confidence threshold, IoU threshold) in the main function.
• Input Data: Update the file_path variable in the load_and_preprocess_data function to point to your dataset.

Example Output

Generated Candlestick Chart

Detected Pattern

Directory Structure

stock-market-pattern-detection/
├── main.py                                     # Main script
├── pattern_detection.ipynb                     # ipython script
├── README.md                                   # Project documentation
├── XAUUSD_M5.csv                               # Example dataset
├── images/                                     # Generated candlestick charts
└── detected_patterns/                          # Saved detected patterns

Contributing

Contributions are welcome! If you have suggestions or improvements, please open an issue or submit a pull request.