TC2000.com• Download software • Tutorial videos • Subscription & data-feed pricing • Class schedule
TC2000Brokerage.com• New account application • Trading resources • Margin rates • Stock & option commissions
Worden Discussion Forum
Stock and Market Talk
Using Deep Learning Neural Network models for trading
Has any programming savvy gurus here ever tried to utilize ANN models based on known trading experts' trades and strategies to screen/predict trades? If you would like to share, what are the outcomes? what are the shortcomings that are encountered while trying to train the model based on trade data and other potential catalytical data?
I've tested various NN models and tools before. I don't personally know of any that tries to use "expert trader trades" as a model to train with or against, but I have no doubt that someone or some organization is using such data. The well-funded quants and algos are unifying such data and far, far, far beyond. I have heard about projects from insiders on those projects that make me question whether anyone should trade at the retail level, such is the extent of the data processing edge.
I don't personally think NN is the area of central interest for such data manipulation these days. NN has had a checkered past in forecasting price data and more interest is in ML or DL now. This could be due to NN's roots in modeling distributions that were typically unlike price anomalies in finance but were more central to genetic modeling, but I'm certain that this problem was examined years ago and explored thoroughly. That is to say, some of the positive zeitgest around NN for finance has been tempered, and that muting effect has been accelerated by the serious, aggressive and intense implementatio of ML and DL across all data.
Data abounds, including "expert trade data", the latter even being discoverable in not so obvious places. Separately, I do think time/sales data exposes presence and activity of the "smart money", and so from a serious financial data scientist's point of view, it would be not distinct from "expert trades" exept that it is probably better since it is happening in near realtime.
A google search of: deep learning technical analysis yields this TensorFlow video
The code is avilable for those looking the explore deep learning via TensorFlow
The paper is measuring the absolute value of the price difference of price vs. the prediction divided by the price. It isn't developing a trading plan and measuring profits which I think would be more relevant.