Performance analysis of study material recommendation system to reduce dropout in online learning using optimal behavior prediction cluster and online poll bot

ABSTRACT

Students are termed “multitaskers,” and it is likely that they easily fall prey to other subjects or topics that most interest them. They occasionally took heed or gave close and thoughtful attention to the lectures they were on. In the current educational system, our young generations receive materials from their leftovers, and their constant behavioral classification has decided the material to learn. The rate at which many students gave up on their studies was predominantly higher in online classroom than in offline classroom due to the lack of direct interaction between the students and teachers. To eradicate this and to make online classroom an effective one, the proposed model can be put forth in each class to predict the student's behavior based on their keen interests. The model predicts and recommends their live session-wise apt course materials to learn. This model uses machine learning generic algorithms and the chi-square test to analyze their manners. The intelligent Online Poll Bot (OPB) acts as a teacher in this virtual learning environment by engaging in live interactions during class time. It is developed using GAN and the IBM Watson Framework. This paper analyzes the time complexity and accuracy of the developed poll bot, and 96.82% accuracy was achieved with the proposed GAN-based poll bot. Students can be categorized according to their learning behavior by using the Optimal Behavior Prediction Cluster (OBPC). These OBPCs will let know the number of clusters at the beginning of the process itself. According to the model, the study materials are preferred based on the students’ performance in each class. In online learning environments, the Live Behavior Analysis (LBA) method using the proposed OBPC and OPB can create interactive learning environments and deliver behavior-based study materials to learners, thus reducing dropout rates. The proposed experiments show that the accuracy of the OBPC-based system is 97.43%, and LBA produces 96.52% accuracy, 95.13% F-Score, 97.13% recall, and 96.14% precision compared to existing approaches. This technology will reduce the number of dropouts and can effectively predict the behavior of all students in the virtual environment where they are placed.

KEYWORDS:

• Students behaviors
• online learning
• optimal behavior prediction cluster
• online poll bot
• dropout
• live behavior analysis

Disclosure statement

No potential conflict of interest was reported by the author(s).