Pengkelasan kompetensi bidang kepakaran berasaskan rangkaian neural dalam sistem pengurusan kompetensi

Abstract

The expertise of a workforce is closely related to the level of competency and other factors such as knowledge, experiences, skills, attitudes and et cetera. The problems of competency management and the methods to determine competency level of a workforce always engender an issue. As for example, what is the best method or system to determine the weightage for each criterion in evaluating a certain level of competency? Among the method to solve the problem is to use neural network. But the accuracy of the network depends on the activation function used. Thus, this project aimed to compare the accuracy of several activation functions for classifying competency level of a workforce's expertise. The comparison of activation function in this study is done by using back propagation neural network and by calculating the percentage of the accuracy of the neural network as compared to the target data samples. This study reveals that the logarithmic-exponential activation function gives a better result as compared to the logistic activation function. The activation function is then used in developing a prototype module To Classify Expertise Level for Competency Management System using The Unified Software Development approach, which apply the techniques of Unified Modeling Language (UML). The prototype module has been used as an efficient and fairly method in term of getting weightage for each criterion in determining the competency level of a researcher's expertise at the Malaysian Institute for Nuclear Technology Research (MINT).