TradingTensors

A simple trading simulator inspired by the OpenAI framework. The objective of the project is to explore the viability of AI (Supervised/ Reinforcement) algorithms in the financial markets. Models built in Tensorflow. Designed to carry out historical backtesting and live trading with Oanda API.

Experiments

• Base Environment Representation
• Experiment #1: Trading with previous n periods of Open-Open Return (DQN)
• Experiment #2: Trading with Moon Coordinates? (DQN)
• Experiment #3: Trading with A3C

Sources: DQN made simple by Arthur Juliani

Prerequisites

• numpy
• tensorflow
• talib
• pandas
• swisseph (Swiss Ephemeris)

Getting Started

1.Installation:

pip install -e .

2. Enter Oanda API ID and token key in serverconfig.py

3. Best viewed in notebooks

This is created by Henry Bee, intended for Mr Peter's use