hi everybody, I'm hearing alot of talk around integrated AI systems into automated trading systems. Has anyone successfully integrated an AI system into their trading platform that has had proven beneficial impact on their trading system's profitability? Doesn't it fly very much in the face of what we know about market movements (i.e. the long term random nature thereof) to put an AI system to work? Its been a number of years since I've worked with AI systems and perhaps I'm been overly simplictic with regard to what these systems can actually do. All opinions welcome on this one me thinks! cassidpp!
1. Define AI. Can you be a bit specific? If you are talking about "pure NN", everyone around me laughs at it. 2. Whether it be NN, GA or etc., it's a tool and very much viable as an algorithm., used part of a process.
no i'm talking about a neural net in its purist sense. Machine learning. And how exactly does one integrate it into an automated trading platform.
Decision Support, Pattern recognition etc. You can treat a neural net as a specialized indicator for example. How is it integrated into an automated trading platform? Well, you pump the net with market data from your feed and other inputs you desire and the neural net spits out decision making data that you can use to make trades. Suggest you search ET for various threads on this topic e.g. http://www.elitetrader.com/vb/showthread.php?s=&threadid=71366 There are a bunch of published texts on this topic also: http://www.amazon.com/Neural-Networ...1215933?ie=UTF8&s=books&qid=1174833836&sr=8-2 http://www.amazon.com/Neural-Networ...1215933?ie=UTF8&s=books&qid=1174833836&sr=8-1 http://www.amazon.com/Artificial-Ne...1215933?ie=UTF8&s=books&qid=1174833836&sr=8-7 If you are a Java programmer then www.jooneworld.com might be of interest. There are links to other books of interest on that site too.
Machine learning is great help to develop patterns that represent complex relationships in information-saturated financial markets. During analysis NN compares existing stock-trading patterns with previous situations and generates trade signals. In other words, NN searches out linkages between stock performance and variables such as price momentum, MAvg, MACD, and so on. Pertaining to how integrate NN you can select a set of indicators and use their values to feed previously trained NN. As outcome you will get trade signals.