Inspiration
ML Trading
What it does
Assigns labels to features using tables of stock prices and ITF web news sentiment
Features = combination of sentiment, news buzz, news volume
Label = 1 is current price > last price = -1 otherwise at that same time
Four training models are run on the sentiment data to predict the label using the data and if three or more agree on the label value, then we buy/sell that stock depending on the value of label. Otherwise, nothing the signal is to hold.
The sentiment data extracted from ITF and each row has the previous 30 days worth of sentiment data compressed into a string. The decompression is done by a helper function and the trading strategy is run.
How we built it
Using Python3, using Keras and Tensorflow. With datasets being pulled from Quandl and then the model tested on Quantopia.
Challenges we ran into
The preprocessing!
Log in or sign up for Devpost to join the conversation.