Arif, HossainAnough, Ahmed Saquif AlamHossain, Md. TahmidKarim, Kazi RuzlanFaruk, Umar2019-10-292019-10-2920192019-08ID 19341020ID 14101148ID 18241050ID 15141003http://hdl.handle.net/10361/12816Cataloged from PDF version of thesis.Includes bibliographical references (pages 44-48).This thesis is submitted in partial fulfillment of the requirements for the degree of Bachelor of Science in Computer Science, 2019.Recognizing a growing need to accommodate students of varied backgrounds and account for individual di erences in learning curves, this paper re ects on our work to implement a Temporally Coherent Clustering approach in order to detect the most optimized pathway for teaching subjects through an MCQ based learning platform. Standard approaches towards extraction of student activity data typically detect similar behavior patterns and use simple statistical analysis in order to make predictions regarding their result. In reality, this causes high noise in the data that is temporally inconsistent and largely inaccurate. We proposed to work with an evolutionary clustering pipeline that can be applied to learning data that we have collected through our Intelligent Teaching System - and aimed at improving cluster stability over a large data set of student behavior. Initially, we have collected and worked on BCS Examination related data, where our results show improved cluster performance of both students and study material, and achieves stability on organic user data in order to be able to detect behavioral patterns and properties of learning environments. As an end result of this whole research, we have incorporated our work into our ITS, which proactively determines student's knowledge level, and automatically determines the best pathway in order to improve their performance. Overall, it deliberately in uences a students capacity improvement in order to passively enable them to answer harder questions by creating an optimized pathway that recognizes the need for individualized learning curves. Overall, we managed to get an accuracy ratio of around 84%, with a silhouette score of 0.53 against an optimized k value of 5 within our clustering algorithm using k-means.48 pagesenBrac University theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission.Temporally coherent clusteringExam performance predictionResult predictionKnowledge level determinationCluster analysis--Data processingMachine learningThesis