Historical Data
David_Belvedere
Posts: 19
Hi all,
I am looking to implement a machine learning based strategy, but will need some training data. What are my options for getting and using training data in the CQ environment? Does CQ provide me a way to get the previous X days minute bar data for a given stock, for example?
Tagged:
Comments
take a look here: https://app.cloudquant.com/application/#/glossary/133
Ah perfect! Can't believe I missed that - thanks so much!
Hi David,
We also interested in hearing what types of machine learning tools would be helpful to you. Our R&D group is working on tools that will help facilitate and accelerate the machine learning training process as well as providing tools to analyze prediction quality.
Thanks for your question.
Is there an object that can store information across different days. This question is particularly relevant for machine learning - a model might be expensive to train, so being able to run model.fit function once a month, for example, and store the model object in a variable that can accessed by the instances of strategy for subsequent days would be very helpful. Unless, such an object already exists?
Is there a way to get L2 data?
Due to the size of L2 data we do not store it historically.
If you look at Level 2 for a symbols such as SPY, the L1 BID ASK may update a few times a second in the middle of the day but below that the book may be 50 to 100 prices deep on each side, each constantly updating.. That is a staggering amount of data.
Once a user is live trading with us we do have some variables available relating to Level 2, but they are only in a LIVE environment (forward testing and live trading only). The variables that are available give the number of shares available down to a set price and max/min price down to a set number of shares.