I was wondering if machine learning models can be applied in HFT. As far as I know, in HFT the average round trip time is about 100 microseconds and traders would get no price advantage if they are slightly slower than others. So we usually employ some simple trading strategies (e.g. event trading, statistical arbitrage) in HFT since they are computationally fast. I think machine learning models, especially neural networks are relatively time-consuming so how could they be used in HFT efficiently? Are there any applications? Thank you!
Hello,
If we use training method like rolling-window to train the model once the new price data is released, then the problem you mentioned is difficult to overcome for high frequency trading. Another way to do is that we train the model on preivous data and just use the model to predict, than the speed is much faster than we expect.
Got it. Thank you so much!