Class Ocjknn

One Class JKNN classifier based on: Khan, Shehroz Saeed. "Kernels for one-class nearest neighbour classification and comparison of chemical spectral data." College of Engineering and Informatics, National University of Ireland (2010). https://cs.uwaterloo.ca/~s255khan/files/Kernels_for_One-Class_Nearest_Neighbour_Classification_and_Comparison_of_Chemical_Spectral_Data-libre.pdf

Inheritance

System.Object

Ocjknn

Namespace: SigStat.Common.Algorithms.Classifiers

Assembly: SigStat.Common.dll

Syntax

public static class Ocjknn : object

Methods

Test<T>(T, IEnumerable<T>, Int32, Int32, Double, Func<T, T, Double>)

Step 1: find the j nearest neighbors of testItem in the set of targetItems. Step 2: for each neighbor, if (distance from test) / (average distance from k nearest neighbors) < threshold accept++ Steo 3: return accept / j

Declaration

public static double Test<T>(T testItem, IEnumerable<T> targetItems, int j, int k, double threshold, Func<T, T, double> distanceFunction)

Parameters

Type	Name	Description
T	testItem	The item, that we want to classify
IEnumerable<T>	targetItems	Items belonging to the target class
System.Int32	j	See algorithm description for details
System.Int32	k	See algorithm description for details
System.Double	threshold	See algorithm description for details
Func<T, T, System.Double>	distanceFunction	Calculates the distance between two items of type T

Returns

Type	Description
System.Double	If the result is 0.5 or greater, then testItem should be accepted as a member of target class

Type Parameters

Name	Description
T	Item type (typically a vector or a label, that distanceFunction can work with)