Quantopian's community platform is shutting down. Please read this post for more information and download your code.
Back to Community
Neural Network with many factors

Good Day everybody,

I am considering to write an algorithm that feeds a lot of data (I have made a list of about 50 factors that might influence stock prices) into a neural network. Because thos seems like a whole lot of work to me I wanted to know whether this has been done before, does anybody know about any posts about this topic and if yes could you post them in the comments?

Greetings
Jakob

3 responses

Hi Jakob,

Are you planning to do this within Quantopian's framework? If you are, then you might encounter numerous problems. For one, I do not know of any open sourced neural network python library that is whitelisted by Q. Secondly, since there are compute time limitations, compute intensive routines such as NN with 50 factors will most likely give you timeout errors. Having said that, if you have a reliable source of price, fundamental or alternative data, I suggest you do your planned neural net outside of Q using your choice of NN library or even a third party software and use Q's new Self Serve Data service to upload your NN predictions for further backtest and analysis. Hope this helps.

Hi James,

thank you for your quick answer, I think you're making a few good points here, I have already somewhat prepared the data and a Framework on my PC and will do most of the Work there. I'll keep you guys posted.

Hello Jakob,

Article titled “Time series and neural network forecasts of daily stock prices” by K.C. Tseng, Ojoung Kwon and Luna C. Tjung might be interesting reading material for you. There's a lot of interesting references in that article too.