Hi, You post reminded me of the following paper Pattern discovery of fuzzy time series for financial prediction Chiung-Hon Leon Lee; Alan Liu; Wen-Sung Chen; IEEE Transactions on Knowledge and Data Engineering Volume 18, Issue 5, May 2006 Page(s):613 - 625 Tell me if you want the pdf. Give some specific key words and I can sdearch the IEEE and ScienceDirect.com if you want. BHC
APS software by Michael Harris, searches through the data whether index, stock etc and picks out those relevent to making profits. One can use the program in Metastock and some others. I found the program worked ( made profits ) but there were big drawdowns sometimes. The program searches for highs and lows etc and isn't neural net.
The journey is more important than the goal, my friend. If it were as simple as using NN in the markets, then everyone would do it. So obviously, it isn't. However, as a Computer Science major and an engineer at heart, I enjoy reading and applying all areas of knowledge, and I think that machine learning can be effectively used in money management and portfolio allocation. Utilizing NNs to identify chart patterns is only a small picture in my further studies of artificial intelligence and its application to financial markets.
Hi Humpy, I second that, I have been using APS for a couple of years with wealth-lab and works very well. I had drawdowns in the beginning but I tuned my stops to limit loses and my equity curve is a lot smoother. The company says APS is " a data mining application" and thus not subject to the training issues NNs have: http://www.tradingpatterns.com/Software/apsinfo/apsinfo.html I think NNs do not offer the prceision necessary for swing/position trading. I am planning of trying one such program in the near future to see what it can do. Have you tried APS with spot Forex data? Any good results? Alex
NNs are great if you wish to perform vol interpolation (GARCH) in a Ritchken-Trevor lattice, other than that NNs are so 90's. Just a hint if you are interested in this sheah: SVMs.
SVMs were next on the list, though I have FAR less experience with them than I do NNs. That journey will take me quite a bit longer. Any ideas which rabbit holes I should start looking in, or should I just go for the deepest one I can find? Thanks -C
Plenty of stuff here: www.kernel-machines.org and with more relevance to finance here: http://www.svms.org/finance/
SVMs is a very fancy way of using dummy variables in your usual linear regression, something that leads to excessive optimization. I think you'll do better using a SMA optimized crossover. Just IMO. Alex