Convolutional-neural-network - Chart Pattern Recognition

Description

• Chart Pattern Image Recognition - Identify chart patterns using Convolutional Neural Networks

Model Type

• Convolutional Neural Network 06-10-2022

Version - V1.0 Beta

Goals:

• Build a chart pattern image recognition model and evaluate if chart patterns are useful to predict future prices in the market

Key Notes and Insights

• At this time the model is only able to classify the charts as bullish or bearish
• You should add images to the dataset if you wish to improve the model

Version Updates

• Live test added
• Live Classification added
• Candlestick chart added
• Importing pickle data
• Add layers
• Fit the model
• Test the model using your own Images

Future implementations

• Plot multiple charts
• Add more data to the dataset
• Add Automated Download dataset from google images
• Add more CATEGORIES such as Flags, Pennant, Cup and handle and so on.
• Add Auto Test Image from any Chart
• Plot Multiple Chart Symbols and Multiple Chart time frames
• Store prediction
• Store real market direction
• compare with different targets
• run loop to get best parameters and results
• Set Cross validation to be able to save parameters
• Set auto run
• Evaluate Results on multiple charts
• Save correct predictions and add it to training dataset (Auto Feed)
• Identify potential targets
• Add Break out as bullish with target
• Identify specific chart patterns such as breakouts
• Set Screener
• If there is any patterns useful it should identify and classify it

Author - Luiz Gabriel Bongiolo

References

• Harrison Kinsley @Sentdex - Deep Learning with Python, TensorFlow, and Keras tutorial - https://pythonprogramming.net/introduction-deep-learning-python-tensorflow-keras/
• Shaan Shah - Identifying Candlestick Patterns using Deep Learning - https://towardsdatascience.com/identifying-candlestick-patterns-using-deep-learning-b7d706726874
• https://www.tensorflow.org/api_docs/python/tf/all_symbols
• https://keras.io/api/layers/