Methodology for generating early warnings in assessments of analog electronics competencies by using machine-learning models

Abstract

This study presents an early warning alert system based on LSTM (long short-term memory) networks that evaluates the knowledge of students with respect to thematic units corresponding to an analog electronics competency. A database is created with student answers to a previously designed test. The database consists of 60 questions that assess an electronic analog competency. The results show that the LSTM-based methodology has a greater than 80% efficiency in determining student deficiencies. In addition, it has a greater than 90% accuracy in predicting answers to questions of higher difficulty, if given the answers to simpler questions in a lower stage. It is concluded that the applied LSTM-based methodology is reliable due to its high precision in generating early warnings and in predicting student answers in the future.