How to build a decision tree?

Question

Mathematics

The question is for those who are familiar, or even at all - making friends with logic and mathematics better than me, well, or simply to educated people, of whom there are plenty on Habré.

Example:

Человек за рулем автомобиля пересекает перекресток. Мы ставим перед собой задачу о том, как классифицировать данную ситуацию, а перед водителем - задачу соблюдения им правила и регламентов, называемые "правилами дорожного движения".

Based on a certain amount of knowledge of these traffic rules, we will try to generate a range of questions, the answers to which will allow us to interpret the situation in generally accepted terminology, draw the right conclusions about what is happening and, further, build a tree of tasks and decisions.

The first picture is a draft of our ideas:

Next, we understand that there are rules that are more important, such as having or, save us from this, not having a driver's license to drive a vehicle and there are secondary rules.

The next iteration of reflection, we accept that the sobriety of the driver is either more important than having a license, or secondary, but the scheme is already changing and taking on the following, more “logical” form:

So here is the question itself. What kind of science or its application governs the construction of true/false decision trees and what is the name of this approach to setting problems, as well as the name of that science or a number of sciences that allow you to build a hierarchy of ideas and check this tree for some correctness?

The above reflections are an example of "on the knee". Tasks are more complicated by an order of magnitude, there can be more than two YES/NO connections between each node, there are more and less priority questions, etc., but I need to build some scheme, classification. And not from the top, calling the range of tasks "traffic rules" (rules of the road). The very general idea, the name of this tree, should appear as such a tree is being compiled. But I do not have a conceptual language, although I am sure that the bicycle was invented long ago.

What are your guesses? What am I looking for? In what sciences do such a "theory of classifications" work? Who has a similar toolkit?

Tags are my hypotheses in this regard. If there was an answer, I wouldn't ask the question.)

Answer 1

It seems to me that this is something from the field of Prolog.

Answer 2

Check out a little theory here . Also, building trees is one of the features of the data mining packages ( example ).

Answer 3

You can look at the "Facet Classification" (colon classification, Ranganathan classification). There, some branches may belong to others, but it does not quite fit into the concept of a tree, as I understand it, more than a honeycomb or moss.

Answer 4

There is a very user-friendly tool RapidMiner
rapid-i.com/content/view/26/84/lang ,en/

Answer 5

Here is a good overview of free datamining packages blog.samibadawi.com/2010/06/orange-r-rapidminer-statistica-and-weka.html should be translated into Habré.

Answer 6

The science is called "Systems Constraint Theory" by Eliyahu Goldratt.
And for the implementation of Lisp to help you.

Answer 7

Here is a great article:
habrahabr.ru/post/171759