Anyone interesting in data analyse?

Sorry for delay.
The algorithm itself works quite simply.
Imagine that we have data represented as a set of vectors. Each new vector is compared with the whole existing ‘forest’ of vectors and becomes a part of the forest itself. In this case, each individual vector can be restored back without loss. As a result of the work, we get a list of all parts (a set of elements in the vector) that we already have in the forest with link to other vectors that have the same part.

For example, if we take a data set with handwritten figures. Then as the result of the algorithm, we get the sets that will be common to all digits, the sets that will be common only to, for example, the digit 5, and these set will play a key role in its recognition.

Unlike neural networks, we can pull out and look at the composition of each set and assign a marker to it. The number of markers is not limited. Markers can be either specific signs or vectors as a whole.