An Immune Algorithm with an Evolutionary Scheme for Component Selection for the kNN Method

We introduce an immune algorithm (IA) that for the generation of a cell for a new (immune candidate) cell group uses an evolutionary scheme that makes the cell inherit receptors from more than two other cells. This IA is used to find combinations of features (components) that could improve the accuracy of a kNN (k Nearest Neighbor) when it is used for diagnosis or prognosis. We evaluated our approach with five medical data sets and found that it effectively helps to improve the accuracy of kNN in more than 2%.