Developing a Computer-Mediated Communication Competence Forecasting Model Based on Learning Behavior Features

Abstract

This study aims to develop a computer-mediated communication (CMC) competence forecasting model (CMCCFM) based on five well-known machine learning schemes－ linear regression algorithm, M5P algorithm, sequence minimum optimization regression algorithm, convolutional neural network algorithm, and multi-layer perceptron algorithm and learning behavior features collected by a micro-behavior recorder from the learners who used a web-based collaborative problem-based learning (WCPBL) system to perform a problem-solving learning activity. To explore the big data generated from a huge amount of learners’ micro-behaviors as the user behavior features for establishing a good CMCCFM, this study developed the learning micro-behavior classification structure according to the collected data features and the concepts of CMC competence. Additionally, an effective method for constructing a CMCCFM with high correctness and stableness was proposed and examined its forecasting effectiveness. The effects of learning situations on the prediction accuracy of CMCCFM were also explored. Analytical results show that the CMCCFM developed by the five considered machine learning schemes all have good prediction performance to some degree. Among the five considered machine learning schemes, the sequence minimum optimization regression algorithm with the model complexity parameter set to 0.1 has the highest prediction accuracy and stability. The key features that influence the prediction accuracy most are “communication behavior” and “communication objective.” Moreover, the development of the CMCCFM must consider the learning situations and learners’ traits, such as discussion trait and familiar degree with the problem-solving subject, because those would affect the prediction accuracy of the developed CMCCFM.