FSV versus human subjective data evaluation; an informal survey

Abstract

Feature Selection Validation has now become an IEEE Standard, and is gaining popularity in the Computational Electromagnetic Community. This paper presents an informal and admittedly unscientific survey of various professionals' opinion of the match of models and measurements. Four data sets were presented to individuals who were asked to give a global value, using the FSV Validation Rating Scale. This was compared to the average Global Data Measure for each data set. It is shown that over a sample of 28 people the average Global Data Measure did indeed agree with that calculated by FSV for two of our sample sets, and may have agreed with the other two under more rigid circumstances. Data was taken as functions of engineering background and levels of experience in the field of engineering in general, and electrical engineering specifically. This admittedly anecdotal data demonstrates that FSV can be a valuable tool, but human experience is key to understanding the use of any data set, and context of its use, as always, remains the most important aspect of evaluating results of the process.