intelligentics

kerosen · #1 (**permalink**) 05-11-2009, 07:37 PM

in order to keep the thread "Artificial Intelligence" clean and not to have other kind of discussion, here I will keep you informed about the Intelligentics project.

who didn't read my first post about this project, could find it HERE. the system rough blue print is in the attachment (again for those who didn't saw it on my first post.

who wonder from what it comes Intelligentics= Intelligent + Information + Creation + Science

and now back to the project.

Data Base

Data Base: will be PosgreSQL and it should store the following(future inputs to store is subject to discuss):
All major Indexes
All major commodities
Economic Calendar
All major currency ticks

The way those data will be collected must not under any circumstances to be broken data (e.g. blackout power or accidental internet connection interrupt)

Is very important to be this way having in view that for AI (artificial intelligence) DM (Data mining) and ML (Machine Learning) modules, the inputs should be as accurate as it can.

Here an external solution like having a script on a 99,9% online server who will collect and record all the data required in a file and the file to be send it to the DB (Data Base) would be a solution. Any others solutions are welcome. or when is done this part, put the module over a few servers on different locations with different ISP and to be synchronized

some tasks to be done, are:

Implement Communication Thread

Introduce JAXB

Configure and Add HIbernate and mappings

DB persistence layer

Investigate how to control Replay Thread

Implement Replay

Test

we use Java

as I said before, we are short of colleagues is hard work and requires good programming skills and time

any comments and help is more than welcome

kerosen · #2 (**permalink**) 05-11-2009, 10:53 PM

We consider that as long time the movement of the price depend on more factors, first is to be analyzed the impact of those factors(e.g. Fundamental analysis , Technical analysis, Economic calendar) in correlation with the pair to be study and only after that analysis is done to start and extract some basics conditions.
We all know that if we discuss about economic calendar, very important factor is the forecast (which at the time for release of that –important-announcement the forecast is already included in the price) as important is release vs forecast.

Trying to apply AI over the capital market should means some strictly rules to follow such as:

Each currency has it’s own characteristics
One pair of currency, has both currencies characteristics
One pattern(e.g H&S) does not have the same impact/accuracy on all currencies pairs
The genetic program should extract the trading rules(long, short or range)
Each currency should be analyzed separately and together with it’s pair
Not always, if a long position isn’t to be taken means short one (it could mean sideways-range)
Fibonacci levels should be taken in consideration.
And so on

We should consider that being a dynamic market with dynamic rules small to medium time frames would be ok to investigate

kerosen · #3 (**permalink**) 05-12-2009, 11:58 AM

here are some shoots about the work we already have done it

crawler for major indexes and the economic calendar

and the DDE application

also we have finished the Technical Indicators module

we are still short to programmers over AI

any help would be great as ideas too

kerosen · #4 (**permalink**) 05-12-2009, 12:01 PM

this is the Technical Indicators module

kerosen · #5 (**permalink**) 05-15-2009, 08:40 AM

Ok… we already start to work on the AI module although this is not our first priority(here you have an attachment over our AI module work). Our first priority is to get done the communication way between the modules and start to gather data into our data base, to put the gathering system on multiple servers, and synchronize them in order the data to be as accurate as it can be.
Beside this, one of the most important thing, as I said it before, is how the AI module and Data Mining module should be used because in our opinion is not at all enough just to take some Time Series and throw it into any kind of AI technique and say “this is The forecast”.

If you look at “Swarm Intelligence in Data Mining” at first chapter you will understand for sure why collecting data is important and why we don’t want to take just simple Time Series. On this book you have also a schema about how is/should to be apply.
So, coming back at my example about a soccer team with high budget and top players (here I call soccer team the AI techniques) and a soccer team with modest budget but a great coach (here I call great coach the strategy to be apply in order to get your AI techniques to work at high percentage), I would prefer the second one.
I don’t say AI techniques doesn’t matter(it matter a lot) but the strategy you use those AI techniques, is above.

We’ll keep you up to date

WE ARE STILL SHORT TO COLLEAGUES

Any ideas or/and suggestions, are more than welcome.