Satisfaction Assessment for Biomedical Engineering Students Using Stacked Neural Networks
Satisfaction Assessment for Biomedical Engineering Students Using Stacked Neural Networks

Summary/Abstract: In literature, there can be found many studies that measure the level of satisfaction in the academic level (University), in all university all over the world. The method of collecting information is based on forms and usually include issues related both to the quality of the education act in itself, but also to the aspects related to the material base of the University, psychological and social aspects, the social perception of the University and sometimes to the further insertion in the labour market. Most studies use mathematical regression as a tool, but there are also works that go for fuzzy or neuro-fuzzy systems. The last two have proved to be very efficient and this suggests the possibility that the merger of nonlinear data can sometimes be a better indicator of the level of satisfaction than the one given by linear regression. We must take note of the fact that, sometimes, a high level of student satisfaction does not necessarily involve a high level of that student’s performance. As a result, we will treat these two goals separately and this paper will focus only on the evaluation of students at the Faculty of Biomedical Engineering, using a tool for measurement and data fusion of stacked neural networks. Bioengineering represents a confluence area of engineering, medical and fundamental sciences. It is difficult to deal with the evaluation technique as a unitary field, since in the students’ curriculum there can also be found, for example, the electric and bio-chemistry fields. As a result, we propose the use of the questionnaire methodology, based on 5 levels of satisfaction, and 4 chapters of interest (Engineering Sciences, Medical Sciences, Fundamental Sciences, and Social Issues) that are weighted in four multilayer neural networks. The results of these neural networks merge into a function of regression, creating stacked neural networks which produce a global assessment of student’s level of satisfaction. The evaluation group was comprised of a number of 127 students in their final year, from three successive generations, about 90% of graduate students every year. Results were employed in the statistical percentage of confidence level of 95%, considered to be acceptable, taking into account the diversity of perceptions on the level of an interdisciplinary faculty.

• Details
• Contents