Monitoring online learners' performance based on learning progress prediction

Abstract

The teacher needs to know the learner's progress so that he/she can improve her/his teaching strategy. The initial challenges faced by the teachers are including the need to monitor each learner's progress so that their needs and difficulties can be recognized as a guide for enhancing learning strategies. One way to solve these problems is to predict student's progress. Mathematics in elementary school is one of the subjects with a high failure rate of learners. In this study, the progress of learners was predicted using Random Forest based on classification using similarity characteristics of learners obtained from the results of previous formative assessment. The random forest algorithm was used because it can classify data that has incomplete attributes, which are usually contained in learner characteristics. Prediction models are built based on data of assessment results from 2 math classes with 46 students in elementary school. The resulting model performance will be measured using accuracy and recall because False Positive is better than False Negative so learning progress will always be improved. The results show that the Random Forest algorithm can create a learning progress prediction model with an accuracy of 90% in training data and 93% in testing data.