Optimal Estimation of Penalty Value for On Line Multiple Choice Questions Using Simulation of Neural Networks and Virtual Students´ Testing

This paper provides interesting findings for modeling of a challenging and critical pedagogical issue namely online learning assessment of Multiple Choice Questions (MCQs) analysis and evaluation. More precisely, in fulfillment of that issue´s objective, this work suggests using a realistic Artificial Neural Network (ANN) model. That, explicitly, characterized by two learning paradigms: supervised learning (with a teacher), and autonomous (selforganized) learning. Furthermore, a computer learning assessment package used for online testing exams adopting a group of virtual 500 students . Herein, a special attention has been paid in order to search for optimal estimated penalty value. In case of multiple erroneous (incorrect) selected answers for random twenty questions submitted by any arbitrary virtual student member out of 500 virtual students. Interestingly, obtained results in case of testing two penalty values (zero & one third) shown to have bell shape close similar to Gaussian statistical distribution. Furthermore, these results have become in agreement with learning achievement results, after running of simulation program of adopted realistic ANN model considering different learning rate values.