Development of analytics tools for e-learning

• Main Author: Vu, Anh Vinh
• Other Authors: Chua Hock Chuan
• Format: Final Year Project
• Language: English
• Published: 2017
• Subjects: DRNTU::Engineering::Electrical and electronic engineering
• Online Access: http://hdl.handle.net/10356/71181
• Institution: Nanyang Technological University

The ubiquity of Learning Analytics as an alternative instrument to traditional education method has surged remarkably ever since the blooming of information technology. The term “Learning Analytics” describe the adoption of various data mining or machine learning techniques in educational field. Plenty of learners and instructors’ data, course contents and feedbacks are used to discover knowledge, insights or patterns of the interested subjects and topics. Popular Learning Management System (LMS) has already integrated this feature but approached differently to diverse targets, some examples are SNAPP or LOCO-Analysis. At Nanyang Technological University Singapore, the current situation requires a “smart” learning system that not only supports operational function of traditional LMS, but also possesses the capability to conjecture student performance, through their constantly changing tracks and study results, hence identify in-need subjects prior to examination. In this final year project, the main objective is to develop data analytic model and to construct standard framework for transforming collected students’ data in a particular NTU/EEE/IEM course into useful information. Subsequently, the next decision made by the professors and course coordinators will be influenced by this information. For empirical system needed to build, the primary purpose is to predict student final exam scores of a subject as soon as we obtained latest continuous assessment results during the academic timeline. Hence, early warnings can be given to students with identification of strengths and weaknesses. Through various stages of data analytics, we have successfully constructed our final prediction model and examined our methodologies proposed. The process includes data pre-processing, models selection and parameters tuning, feature importance and error analysis. The predictor has been built upon different input-output scenarios to answer different questions relating to student performance. Based on the derivation during our implementation, a general procedure with emphasis on high flexibility and robustness was designed to apply to similar NTU courses in the future.