llsourcell / q-learning-for-trading Goto Github PK

Python 100.00%

q-learning-for-trading's Introduction

Overview

This is the code for this video on Youtube by Siraj Raval on Q Learning for Trading as part of the Move 37 course at School of AI. Credits for this code go to ShuaiW.

Dependencies

Python 2.7. To install all the libraries, run pip install -r requirements.txt

Table of content

agent.py: a Deep Q learning agent
envs.py: a simple 3-stock trading environment
model.py: a multi-layer perceptron as the function approximator
utils.py: some utility functions
run.py: train/test logic
requirement.txt: all dependencies
data/: 3 csv files with IBM, MSFT, and QCOM stock prices from Jan 3rd, 2000 to Dec 27, 2017 (5629 days). The data was retrieved using Alpha Vantage API

How to run

To train a Deep Q agent, run python run.py --mode train. There are other parameters and I encourage you look at the run.py script. After training, a trained model as well as the portfolio value history at episode end would be saved to disk.

To test the model performance, run python run.py --mode test --weights <trained_model>, where <trained_model> points to the local model weights file. Test data portfolio value history at episode end would be saved to disk.

q-learning-for-trading's People

Contributors

Stargazers

Watchers

Forkers

jshuadvd dinghe doyley91 2series gabeochieng zippety jabogithub estoilkov devbrandon subhodeep assurecapitaladvisors stjordanis wanjinchang jwmneu shixguang jdc08161063 jibols roc5cor 2877992943 demula hosungjung01 kavymp qomposer nikokvcs fathisehla amrahstija hongpingwei muddy91 advaithd dahue githubcucu shaswatac2 greg2paris dexterchan maxgillham raphaelraia kotstantinovskiy dmdv dutchwebs mzahran001 stucking tai-o drwtham x-crypt bigplaylab mucoucah jaminalder kiran-kowda yunxileo paolominguzzi challengeyy deeplearningboy tylee33 younglou1995 nisheethjaiswal rbramwell cysheen jhonnig jrpch raghavb harishmalan waltsworld gostevehsieh oussamaazeddine henryurlo lossoh xiaogaogaoxiao abraham314 fikritrader peterhwang2010 scho15 mayuoak jolyjoly sissixxz mrcolo huberthalim ee304041 wilowisp ernestcr qarchli ashwin2201 joestanton wagnernoise anuragtomar2510 antoinemath terrencehenry datamining4finance cmshop jackzhch hmavrodiev aemorales sangkny strong-roots-capital sahanduiuc tibaes todun abhi-h erichuang2013 egoliuk ljferrer

q-learning-for-trading's Issues

More Fatures

Anyone a hint how i can add more features then just the close price? i am not sure about the prediction values? How does that work?

Thanks

You can't buy and sell at closing price, it's not realistic

If you are training (or backtesting) using closing price, then you have to use the open or close from the next day (or bar). Otherwise it's not realistic at all.

A better example

I fixed the issues in the environment and made a nice Jupyter Notebook: https://github.com/FutureSharks/ml-finance/tree/master/examples/1-deep-q-learning-simulated

This example uses simulated prices to prove the model is learning something.

TypeError: 'map' object is not subscriptable

please help me to pass this problem , i use /run.py --mode train
@llSourcell @IsaacPatole @AntoineMath @S0AndS0

IndexError: tuple index out of range

I have met an error as follows:

File "C:\Users\User\Desktop\research_project\run.py", line 63, in
next_state, reward, done, info = env._step(action)
File "C:\Users\User\Desktop\research_project\env.py", line 73, in _step
self._trade(action)
File "C:\Users\User\Desktop\research_project\env.py", line 97, in _trade
action_vec = action_combo[action]
IndexError: tuple index out of range

May I know how to fix it?

TypeError: Error converting shape to a TensorShape: int() argument must be a string, a bytes-like object or a number, not 'tuple'

Canceled

Error during executing run.py file

I am getting the following exception:

None
Traceback (most recent call last):
File "run.py", line 56, in
next_state, reward, done, info = env.step(action)
File "C:\Users\asgha\Anaconda3\envs\tensorflow\lib\site-packages\gym\core.py", line 96, in step
return self._step(action)
File "C:\Users\asgha\OneDrive\Documents\AI\ML\Forex_trading\Q-Learning-for-Trading-master\envs.py", line 68, in _step
self._trade(action)
File "C:\Users\asgha\OneDrive\Documents\AI\ML\Forex_trading\Q-Learning-for-Trading-master\envs.py", line 91, in _trade
action_vec = action_combo[action]
TypeError: 'map' object is not subscriptable

Give credit to the creator you copied the code from

Your content: https://github.com/llSourcell/Q-Learning-for-Trading

Original content: https://github.com/ShuaiW/teach-machine-to-trade (@shuaiw)

It's such a shame that you are running an entire curriculum using code lifted from other people's repositories. The least you could do is give credit to them.

I know it's not a breach of Github's T&S but I expect you to have a decent code of ethics. You don't need to be original and expert but at least be nice and humble. A big chunk of your videos is about finding motivation and being smart about learning. Surely you don't want to come across as a hypocrite.

NotImplementedError by gym

I am getting the following error which executing run.py
state = env.reset()
File "C:\Users\hp\AppData\Local\Programs\Python\Python36\lib\site-packages\gym\core.py", line 69, in reset
raise NotImplementedError
NotImplementedError

Some people have suggested to change env.reset() to env._reset() , I tried it as well but nothing works