K neariest neighbours is a self learning application basically used for classification and regression. It can foresee the values of the decision systems (decision system is an array with values, which represents parameters/actions/etc and decisions made for them) which is perfect for AI in games. Thanks to knn you can simulate intelligent behavior of npc or simulate the stock exchange values and many more purposes.
Asset is great for learning the knn ideology (school purposes), but it can be easly modified to use in games.
In this asset you can load training system thanks to which one the program is learning the rules, and test system, which checks the correctness of decisions given by program. You can choose the methrics (Manhattan, Euklides and Cannber) and a number of neariest neighbours.
As a result an application return a prediction matrix which tells you what kind of decisions were found in the system, number of objects of each decision, accuracy, coverage and true possitive rate of made calculations,
For more information about the algorithm structure please visit this site
