Levenberg-Marquardt Algorithm Based Neural Network Model for Predicting Licensure Examination Performance of Civil Engineering Students

Abstract

In the onset of the 4th Industrial revolution, wherein the use of artificial intelligence is mushrooming in research, the Artificial Neural Network (ANN) algorithm has become an advanced tool when it comes to building of performance models. This study utilizes ANN using MATLAB to create a model that can predict the performance of civil engineering students in the Licensure Examination.
Employing educational data mining techniques, the ANN model output could identify the student’s performance in the Licensure Examination for Civil Engineering. The ANN models utilized Feed – Forward Back Propagation and Levenberg-Marquardt algorithm due to its simplicity and wide array of use. The utilization of the samples was distributed into three phases: training, validation and testing phase. Three (3) licensure examination periods were used for the creation of the prediction models from 2011, 2015 and 2018. The basis for the selection of the chosen periods was based on the change in the number of items in the licensure examination from 30 to 100 to 50 items. The input parameters were the student’s academic performance in the different subjects divided into three categories patterned from the licensure examination criteria. The output used in modelling is the respondents’ board examination score. From the data gathered, three (3) models were created for the three (3) civil engineering board exam subject areas..
Higher Education Institutions (HEIs) will be guided in determining the student’s predicted performance and to carry out measures to give priority to the low performers. The identified civil engineering students should be given higher priority during the conduct of major and correlation courses in their terminal year. The early prediction data can help institutions to implement solution to improve the actual performance during licensure examinations. Using the output models and equations, the students can easily identify their predicted licensure examination performance integrating their academic records from the school and likely will give them proper motivation to improve.