Deep Nexus One (DN1) Research Repository for FX Prediction

Project Overview

This repository contains a foreign exchange (FX) trading prediction system developed in 2019. The project implements machine learning prediction models using an LSTM architecture with a regression output. Multiple models run in parallel and send signals to MT4.

Technical Architecture

Core Components

• Prediction Models:

  • Implemented using TensorFlow and Keras
  • Long Short-Term Memory (LSTM) neural network architecture
  • Developed for 6 different currency pairs
  • Prediction frequency: Every 5 minutes

• Backend Framework:

  • Flask web framework for serving models
  • Waitress WSGI server for deployment
  • Oanda Rest API integration for market data and trading execution

Technical Complexity

The project represented a significant technical challenge in:

• Coordinating multiple LSTM neural network models
• Integrating diverse technologies (TensorFlow, Keras, Flask, Waitress)
• Serving multiple prediction models simultaneously
• Executing trades across multiple currency pairs from a platform originally designed for single-pair trading

Key Technical Challenges Solved

1. Multi-Model Management

   • Developed a custom .dll file to manage multiple prediction models and scripts
   • Implemented a timer function to orchestrate script execution across different currency pairs

2. Platform Integration

   • Bridged prediction models with MetaTrader 4 (MT4) trading platform
   • Resolved complex compatibility issues between different programming environments
   • Managed communication protocols (largely involving TCP/IP networking)
   • Implemented asynchronous processing to overcome limitations of MT4

Technology Stack

• Python
• TensorFlow
• Keras
• Flask
• Waitress
• Oanda Rest API
• MetaTrader 4 (mql4)

Deployment Considerations

• Custom .dll file for script and model management
• Periodic execution of prediction models
• Seamless integration between machine learning predictions and trading platform

Limitations and Considerations

• Research project from 2019
• Developed for specific trading strategy and currency pairs
• Requires careful review and potential updates for current market conditions
• MT4 trade logic is built from a genetic algorithm and is sample logic only
• Python 3.5 was used to build this project
• Final production models are proprietary

Disclaimer

This is a research project and should not be considered financial advice. Trading involves significant financial risk.

License

http://www.apache.org/licenses/LICENSE-2.0

Contact

web@deepnexus.com

Repository initialized in 2024, based on research conducted in 2019.