Intelligentics
"Understanding the PAST, and watch the PRESENT, we can look into the FUTURE"
Abstract
Here in this article, we describe the way an Intelligent Evolving System should be build and work for any giving problem who would involve the following procedures: analyzing, learning, supervise, and execute. Toward an Expert System which is build for a very specific problem, an Intelligent Evolving System should be beyond because it can "understand" almost any giving problem.
Whatever the problem, environment or system is to be analyzed, imply four simple, general, questions:
When?, What?, Why? and How?
In first chapter I will try to explain or define those questions address to a an environment or system what is about to be analyzed.
In the second chapter I will describe a framework and the way we should build an Intelligent Evolving System that by it self should "think" and work guided by questions stated above.
The first big issue is that the Intelligent Evolving System in the beginning it doesn't know anything about the environment. So first think to do is to start and "look" in the PAST. The PAST it is recorded in a Database or warehouse data. What every warehouse data or database have in common? Is TIME record. So every sequence or pattern or whatever that WD(warehouse data) or DB(database) has recorded, is or it should be insert into a chronological order or if not, there in any PAST the time dimension it is involved, which means that even if the (IES)Intelligent Evolving System does not know a think what is about to get into it, for sure it knows that there is always a time record. So here we have a first simple rule: TIME
Anything what happened in the PAST it is related to TIME, and this means that our first question it is done (When?)
Starting from a saying "what goes up, goes down" and improved a little, we can state, "what goes up, goes down, side ways or stays put" and this should take us to the DB and discover over the TIME constant "What" moved in those directions. Up, down, sideway or stays put. This should fulfill "What?" question.
"Why?" now this is a difficult one. Taken from human being, usually each one of us has his own opinion when he is questioned about something with "Why….?". and this because everyone has his own perception about that thing regarding "Why" and his perception depends on his motivation, experience, vision, education. When those reasons that influence our perception about "Why…?" are close to others then the perception should be close to others or slightly modified. This statement make us to state that usually when experience, education vision and motivation are close to others, there will be almost same perception about "Why…?"
Having a good understanding of the above questions("when?", "what?", "why?"), it would be much easier to understand "How THIS does work out?"(whatever THIS means) in order to "…watch the PRESENT and look into the FUTURE"
If we understand "when" it was happened, "what" was happened and "why" was that happened, then we can define "how" to watch(supervise/monitor) the PRESENT in order to look into the FUTURE.
Of course all this questions can and must be asked in any direction and level of the application.
Setting up the tools:
In order to get the answers, we have to put "someone" to search for it. This "someone" should be an army of intelligent agents, each one with specific design, for specific task. Out there, including in the open source projects, are very powerful applications who can be used in order to build this army of intelligent agents.
This is like having hired the perfect employers (working 24/7, no wedge, no holydays and so on), as many as we need, to do the search for what we looking for.
What we should look at an intelligent agent when we "hired" it, is its reliability, and speed.
On the intelligent agents world we can find almost anything a employee need, including buyer agents, user agents, monitoring/surveillance agents, data mining agents and so on.
After we decide what kind of intelligent agents should we used and for what tasks, we should set up the Data Mining module. This module should help to "Understanding the PAST…" part, which means this is another milestone in our final goal. The way it select the features(inputs) the way it extract them, is very important. It is important because this would be part of future "education"(remember that perception over a subject depends on education and experience in that subject's domain) of our AI module. A very good "education" and a "big experience" lead to an "expert". The data mining module should it self discover association rules or build decision tree or whatever techniques it uses. Should be let by it self to do it, and also by it self for instance, to check if those rules apply in other particular moments others than it extract for.
Here again, the internet is full of open source data mining applications. We should choose one who is widely well-known, as a reliable one and using various data mining techniques.
This is the module who should take care of the last part "… look into the FUTURE". It contain Artificial Neural Networks and everything related to ANN. It is widely known that ANN have great capability to learn and memorize and are good into forecast, pattern recognition, time series and so on.
The module contain:
ANN builder where ANN are design and deployed automatically.
ANN training and validation module, where by different techniques(supervised or reinforcement learning) ANN are trained and validated
Out there on the internet, are good reliable open source applications for ANN.
And here we have the Intelligent Evolving System. Of course this is a framework which has its details. Of course there are things to talk about it like the way we should train the ANN or by using evolutionary algorithms or genetic programming to optimize parameters like features or ANN and so on but this are details and everyone, as a good "teacher"(for data mining) or a good "coach"(for ANN) should know crystal clear what to look for from such systems.
