KNN method applied to tests of artillery igniters

In the introduction of this article the „k” nearest neighbors method is presented, which can be use both for classification and regression tasks. The process of designing and built the model of the „k” nearest neighbors method was characterized, noting the lack of a generalizing model for the analyzed problem in the method and the possibility of making predictions for new predictor values. Based on the diagnostic laboratory test results, the „k” nearest neighbors model was designed for artillery igniters of the KW-4 type. The predictors and input parameters were specified, on the basis of which the proprietary model was built in the „k” nearest neighbors method. The obtained form of the model was analyzed in terms of the validity of the cross-test and the value of the number „k”. Further models were also built according to the „k” nearest neighbors method for other artillery igniters and for RGM-6 fuses, the results of which were much worse than the KW-4 igniters. Finally, the model was designed according to the input parameters suggested by the Statistica software. The model built in this way turned out to be the optimal model from the point of view of the obtained initial indicators and it should be implemented for use.