Differences between the method of mini-models and the k-nearest neighbors on example of modeling of unemployment rate in Poland

Abstract

The paper presents in a possibly reader-friendly way, in the 2D-space, the method of mini-models, which very well suits for modeling economic dependencies, where frequently a part of explanatory variables influencing the explained variable is not known because lack of data. Experiments realized by authors confirmed superiority of mini-models over such modeling methods as polynomials, GRNN-neural network, and the method of k-nearest neighbors (KNN). Because the method of mini-models is frequently mistaken for the KNN-method the authors explain in the paper the significant difference between the both competitive methods. The indicated difference is also the main reason of superiority of mini-models over the KNN-method. Accuracy of both methods has been compared experimentally on example of modeling unemployment rate in Poland and also on examples of other economic dependencies.