PREDICTING MARKETS WITH DEEP LEARNING MODELS

A vital feature of this model is a complex network structure with two inputs with distinctive layer types in each branch. It has 50 LTSM cells in the first layer, a dropout layer to prevent overfitting, and a few dense layers.  A highlight of this network the linear output activation that allows the model to accurately tune its penultimate weights. 

I also created a simple bot to make buy/sell decisions for the day. The bot predicts the opening value of the stock for the next day, and if it is beyond a threshold amount, it buys the stock. If it is below another threshold amount, it sells the stock.

predicting markets deep learning models.pdf

Download PDF • 357KB