I guess this trick could work in a penny stocks which are mostly bound to circular trading. These kind of stocks have very little or no influence of external real time factors. If at all there is any external factor, a simple trigger might be an anomaly on nupic. But of course, nupic should be able to learn about these kind of stocks on a ms basis. Bringing in social media like twitter mentions, web alerts as an external factor might even add more weight as a "positive anomaly" to take an action on the stock. Do you think this makes any sense ? just my initial thoughts after reading this thread.
On Mon, Dec 15, 2014 at 7:35 PM, Matthew Taylor <[email protected]> wrote: > Sounds interesting. How would one get this data? > --------- > Matt Taylor > OS Community Flag-Bearer > Numenta > > > On Fri, Dec 12, 2014 at 5:57 PM, Michael Davidson <[email protected]> > wrote: > > Daniel Bell <john.mrdaniel.bell@...> writes: > > > >> > >> > >> That is certainly understandable and fair. So this is a practical > > limitation of not having visibility on all of the relevant factors. > >> Could nupic do this if we theoretically did have all the features that > > represent the state of the system? > >> Would a subset of these features, no matter how large, be able to > resolve > > 'reasonable' predictions? > >> > >> > >> > >> On Wed, Dec 3, 2014 at 3:16 PM, Matthew Taylor > > <[email protected]> wrote:Hi Daniel, > >> Can any one human being predict stock market prices with any accuracy? > >> If you think about how many factors actually affect even a single > >> stock price (economy, inflation, weather, time of year, time of day, > >> moods of investors, CEO scandals, other stock prices, I could go on > >> and on...), it would be extremely hard to identify them all, much less > >> isolate them into individual scalar values and feed them into NuPIC. > >> There are just too many unknown factors involved. Even the best human > >> minds can't do it. > >> --------- > >> Matt Taylor > >> OS Community Flag-Bearer > >> Numenta > >> > >> On Tue, Dec 2, 2014 at 5:51 PM, Daniel Bell > >> <john.mrdaniel.bell <at> googlemail.com> wrote: > >> > Hello, > >> > > >> > In one of the talks Jeff Hawkins mentioned that stock market data > cannot be > >> > predicted with numenta. Why is this the case? Is it not an appropriate > >> > problem space? > >> > > >> > My question here really is, what are the limitations and how do we > identify > >> > problem spaces that will work well with numenta and not work well > prior to > >> > an attempts to train/predict? > >> > > >> > Regards, > >> > > >> > Daniel > >> > >> > > > > > > Guys, > > > > I have for sometime wondered if nupic could take any stock's Depth of > Market > > datafeed (Level II and Time&Sales) and learn the patterns of market > > participants by their ID's as they post, change and cancel their bids, > asks, > > sizes at different levels to game the stock price. After several days > > (weeks?) of learning on this high frequency data from the same stock, I > > wondered if nupic would be able to discern a relationship between the > > pattern of action throughout the day of certain key participants it had > > classified as market movers (anomaly detection?) and start to make > > predictions of price along with confidence scores at different time > offsets? > > When you use the timestamped quote changes of the participants at every > > level as price is discovered from ms to ms, would nupic show a better > grasp > > of how these influences collude to shove the price one way or another a > few > > seconds or minutes into the future? > > Macro news forces like economy, inflation, weather, etc. might introduce > > some noise but would be priced in quickly and in any case, would be > > represented by the moves of Market Makers being learned from the stream. > > > > (Sorry for the crummy run-on sentences, I'm in a hurry tonight.) > > > > Michael Davidson > > > > > > > > > > > >
