Bitcoin Trade Data

The goal for this project is 1)to get bitcoin related data to predict bitcoin price. 2)Predict Bitcoin future price 3)Auto-trading based on the prediction(Long Only)

The related data including:


quarterly and weekly bitcoin price from OKEX
asks/bids info and last price from Bitfinex (Will add funding data later)
USDT OTC price from Huobi.pro
Emotion from news and forum (Wallstreet.cn and 8btc , will add reddit and CCN later)(DELETED due to effeciency reason)

Outcomes:

3 predicted outcomes: Next_5,Next_10,Next_15 refer to the bitcoin future price will raise or drop in next 5/10/15 minutes based on our prediction

Machine Learning Model:


Neural Network (Prepare to use deeplearning model to fit the data in the future)

Steps

1) run input_data.py to collect data to SQL Database.

2) https://github.com/benjaminshi02003220/Bitcoin_price_prediction Download this project and run NNC.py to build up the model based on the data in SQL Database

3) Copy the 4 model file from Bitcoin_price_prediction folder to bitcion_trade_data folder. Run prediction.py to 1.predict price 2.Auto-Trading based on the result(To be completed)

You need to install below packages before run input_data.py
pip install requests -- Access to webpage
pip install BeautifulSoup -- Formalize webpage
pip install snownlp -- Analyze emotion of sentenses in Chinese
pip install pyodbc -- connect to SQL Server DB
pip install sklearn -- Machine Learning Model
pip install tensorflow -- install tensorflow

Will update the code from time to time.


------------------------------------------------------------------------------

Updates by Date

Updates 3/6/2018


1. get the news from wallstreet cn(https://wallstreetcn.com/) blockchain channel. Analyze the emotion thru SnowNLP. Input as a feature.
2. get the posts from Chinese biggest BTC forum--8btc cn(https://8btc.com/) blockchain channel. Analyze the emotion by using SnowNLP. Input as a feature. (Will Update and input the emotion from Reddit and CCN in the future)

Updates 3/7/2018


Adding solutions for disconnect conditions

Updates 3/8/2018


1.fixed a bug in bfx.py
2.*Changed the data storage from csv to SQL server database.Better for further machine learning process.

Updates 3/10/2018


Start another project named Bitcoin_price_prediction to predict bitcoin trend.

Updates 3/14/2018


DELETED Emotion from forum due to effeciency reason

Updates 3/14/2018


adding prediction.py (To be Updated)

Updates 3/18/2018


Fixed some bugs and complete the real-time prediction. (prediction.py). Next step is to accomplish the auto-trading based on the prediction result~

Updates 3/28/2018


3 predicted outcomes: Next_5,Next_10,Next_15 refer to the bitcoin future price will raise or drop based on our prediction , 1 means raise ,0 means drop. If two of the three outcomes are 1, we can long bitcoin future and will sell it after 15 minutes/or a certain percentage of profit(to be tested)

Updates 4/4/2018


Adding future trading process(Long and sell after 15 minutes)