A first pass on creating LSTM Neural Networks in MATLAB to make day-to-day stock trading decisions.
This project represents my first attempt to create an automated stock-prediting algorithm utilizing machiene-learning techniques. At the current stage in this project's development it is able to automatically scrape daily stock data from the web, as well as download large quantities of historical stock data very quickly. These data can be processed by a modular LSTM neural network generator.
The eventual hope of this project will be to have a master neural net process the daily predictions made by the stock-specific neural networks to make market decisions with the full scope of the SNP 500 (my current set of tickers that I follow).
While initial results are extremely promising, I am currently limited by the computational power offered on my hardware and am investigating new ways to optimize the algorithm to make the computational times viable for day-to-day trading.
Feel free to email me with any questions/thoughts at [email protected].