The prediction system Pibui
I just updated to Pibui d & rsquo; incorporate a simple voting system (+1/-1)and that & rsquo; d & rsquo a system simpler predictive chaining based on these votes.
I said to myself, so, that I & rsquo; going to explain how it works. In advance, do not expect an AI, c & rsquo; s just random “headed” ^^
The base of this system is the voting. Except that, at the BDD, I backup the vote as being associated with a clip but as being associated a transition between the clip and noted the previous clip (so that the note of the first clip viewed n & rsquo; is ignored). This vote is recorded when retrieving the next clip shows.
To know what the next clip to be displayed, I get the list of transitions + notes where the clip that & rsquo; we have just seen is the “source”. Si un clip “destination” n’a pas eu de note (and therefore n & rsquo; is not in this list) or a negative note, he will have a chance on the total number of clip (including filtering via chain) d & rsquo; be displayed (I do not exclude any clip). Si la note est positive, in addition to the base chance, it will have much more chance of rating data, multiplied by [nombre total de clip]/250 (the site n & rsquo; not very consulted, I think it & rsquo; is a good value, but it can go up if more votes are made).
Then just make a random among 0 and the total number of clip + the total number of additional chance for the clips and take the clip that emerges.
That is roughly the current system. C & rsquo; is a very simple system, based on the random and learns as and as notes are given in clips.
It may change over time but at the moment, it works as described here ^^