A collection of materials, links, and resources for starting the next Tensorflow project.
Topic: Qlearner for stock prediction. Using Python and Tensorflow, and the Poloniex public api.
Goal: Have all the links in one place for documentation and reference.
Project space: https://github.com/Ruckusist/tf_polo
Materials:
- Tutorial/Project for using twitter data for stock prediction.
https://github.com/zhedongzheng/stock-market-prediction-using-deep-learning
Some results: https://zhedongzheng.github.io/ - Siraj is a great source for any work he has previously done with Tensorflow, it will be easy to understand.
https://github.com/llSourcell/How-to-Predict-Stock-Prices-Easily-Demo
https://github.com/llSourcell/predicting_stock_prices - This looks super clean and readable even if it has no comments.
https://github.com/ShubhangDesai/tf-lstm-stock-market - Another clean setup. looks like this guy, started an account then only worked on this for a couple days then abandoned it. Thanks dude!
https://github.com/shawnliucmcm/StockPrediction_RNN - This is a currently in progress project by a University student.
https://github.com/dhingratul/Stock-Price-Prediction - This is another older abandoned project, but this has the Qlearner implementation taken from the Cartpole game from the OpenAI gym tutorial. I have previously tried to implement this in chess and will do another blog about how important this exercise is, and how it can be transferred to solve many different problems.
https://github.com/kimber-chen/Tensorflow-for-stock-prediction
Some further reading on topic: https://github.com/kimber-chen/Tensorflow-for-stock-prediction/blob/master/graph/DistributedTensorFlow%26StockPred.pdf - It looks like this person started with the mnist tutorial from Tensorflow and then started making changes from there, is good with comments.
https://github.com/BanditCat/tfstocks - This is a currently running poloniex bot project, might be good for notes.
https://github.com/BitBotFactory/poloniexlendingbot - Another currently running poloniex project. This guy has a good idea on how to interact with the api to get results to the screen. This project is ongoing for a few months now and thats also a good sign and its in python. So many of these projects are abandoned, there are sad statistics on the subject of how many github repos are abandoned and out of date.
https://github.com/BenjiLee/PoloniexAnalyzer - Last but surely not least, is Harrison. Pythonprogramming.net, github.com/sentdex. This guy made his start doing stock market analysis freelance and selling his dataset/findings for what i would guess is good money. Im going to link a bunch of his material on the subject of python-stock analysis. As you finish reading this. This link is where you should start, then go back to the top and get through the rest of the reading material before starting this project.
https://pythonprogramming.net/machine-learning-python-sklearn-intro/
https://pythonprogramming.net/machine-learning-tutorial-python-introduction/
Starting a new project: Create a github repo, pip install jupyter-notebook, clone that repo to your machine, then just type jupyter-notebook in the starter project and work inside of your favorite web browser.
Future Plans: Once work in the notebook shows promise, we can move the whole package into a curses app for a headless server(AWS) implementation. Using the AWS server instead of my personal machine has a lot of advantages including using their internet connect instead of your own for web api call, downloads, and long-term service.
Contact: Please comment if any of this is skipping steps that anyone new to the subject wants to be covered to catch up. Any questions about progress or work done using these materials please follow the project github repo, or look for additional blog posts. Post any additional material links that might be useful and I'll update the post for the community.
Cheers.