Neural network design and evaluation for classifying library indicators using personal opinion of expert

Abstract

PurposeOver the past decades scientific advances have inspired major technological innovations in academic libraries. Academic libraries have transformed into information centers of new quality, becoming thus an integral part of an academic institution's teaching and research curriculum. Assessing a library's functions and services has become an imperative need. The scope of this study, therefore, is to define a theoretical model for the combination of all individual assessment indicators into one single number‐indicator.Design/methodology/approachThe Monte Carlo technique is suggested in order to deal with the problems of the objectivity expert's decision and the collection of a large number of indicators. The methodology used is articulated into four stages: the normalization of single values; the setting of the indicator weighting according to the expert's opinion; the construction of a well‐fitted neural network; and the training and testing procedure of the neural network.FindingsThe proposed method solves problems related to the normalization of measured data linked with significant relevant properties of the library services evaluation.Originality/valueThe practice of using a neural network in evaluating library services as introduced by this paper is characterized as novel.