Inspiration
What it does
This project attempts to decipher the impact of integrating the GS score on a stock portfolio trading scheme, on a total return and Sharpe ratio. This portfolio systematically allocate trade based on a 8-layer sequential CNN prediction in Keras on top of tensorflow, with or without the score.
How we built it
Starting Position: Equally weighted stocks in the input array, with 10% cash buffer. For example, if using Dow, each position would start out with 3%
Trade Allocation: For each day when there is a predicted 1% gain in the predicted stock value the next day, that position would get a 1% increase in position, and vise versa. If 100% of funds is allocated, no further action is needed.
Transaction cost: 1/100 of 1%
Challenges we ran into
Configuring external data vendor took so long that we did not accomplish all the planned features