O PROBLEMACH STOSOWALNOŚCI MIERNIKÓW SYNTETYCZNYCH DO PORZĄDKOWANIA OBIEKTÓW
ABOUT THE PROBLEMS OF APPLICABILITY OF SYNTHETIC MEASUREMENTS FOR ORDERING OBJECTS

Summary/Abstract: The problem of building a stable synthetic index used to organize the objects described with multiple partial indexes has been and is the subject of the authors' considerations. In our research, we aimed at the construction of a measure that would be insensitive to the input data, the results that would not depend on the method of normalization of the variables, the choice of the distance measure (similarity) and the selection of features. Our experience is consistent with the results of the work of the outstanding American statistician L. Breiman. A single classifier (synthetic measure) may be far from optimal, while combinations of many result in a classifier that is close to optimal and stable. Unfortunately, when "weak" classifiers are used, the combination may result in an even worse classifier. In the work, we presented the consequences of choosing a synthetic indicator for organizing objects described by many features on a practical example of farms in the EU countries covered by the FADN survey.

• Details
• Contents