Use of Subspace Clustering Algorithm for Students' Competency and Subject Knowledge Assessment

Use of Subspace Clustering Algorithm for Students' Competency and Subject Knowledge Assessment

Dimple V. Paul (Department of Computer Science, DM's College and Research Centre, Mapusa, Goa, India) and Chitra S. Nayagam (Indian Institute of Technology (IIT), Farmagudi, India)
Copyright: © 2018 |Pages: 14
DOI: 10.4018/IJKSS.2018040104


Student performance studies are the primary challenge for any course with continuous assessment. The challenge lies in performing validation tests of whether course objectives are being met and also in identifying areas of the course structure that needs improvement. This article identifies whether objectives of the course are being achieved or not, by analyzing the student performance in different courses using competencies as the criteria for assessment. Performance evaluation includes diverse types of competency components such as presentation, assignment, group discussion, etc., along with written examination in order to assess the knowledge of students, as well as their interest in the subject. A PROCLUS algorithm has been chosen for experimentation, as the algorithm identifies similarities among data sets and forms clusters of disjoint sets. The algorithm not only considers random sample points, but also successfully scans entire data sets to identify meaningful dimensions that are needed to form actual clusters. Experimental results have identified the similarities of the students' performance across the subjects that are similar in nature and their competency parameters were also found to be similar. A majority of the students have performed alike in certain subjects that involved practical components or in other ways, similar performance is achieved during the assessment of courses on competencies like presentations skills, group discussions, writing skills, etc., rather than mere theoretical components. This study could help to modify the evaluation and assessment pattern for the theory subjects and/or to fine tune the course structure and objectives of such course, and also to find some alternate techniques to improve the other competencies.
Article Preview

1. Introduction

Students’ academic performance analysis has become an imperative and challenging task for various universities and institutions to ensure the progressive growth of every student. In the conventional education system, academic performance had been monitored only at the end of every semester. The system was unable to give importance to steady and consistent growth of knowledge which is an essential component in education. The end-semester evaluation pattern failed to verify the subject knowledge of a student rather could generally confirm the memory power of the student. In order to overcome the limitation of conventional education system, continuous assessment has been introduced. Continuous evaluation pattern consists of various types of competency components such as presentation, assignment, case-study, quiz, open-book-test, group-discussion etc. along with the written examination. The continuous evaluation pattern facilitates to improve students’ performance in a consistent manner as well as assist to achieve in-depth knowledge in the respective subject (Shovon and Haque, 2012). Knowledge improvement in the specified set of subjects of a course contributes to better employability of the student.

Having large data sets of students’ performance data of continuous assessment, data mining technique has been found efficient to extract useful but hidden information is discussed by authors of Aggarwal and Yu (2000) and Kabakchieva (2013). The information if extracted can strengthen the confidence of students with good academic performance and improve their forthcoming performance to achieve better prospects. Also, students with poor performance get to improve in advance to meet the challenges and requirements of the job market.

Data mining techniques have been used with relational databases to discover unknown patterns, searching for unexpected results and correlations. Clustering which is considered to be the unsupervised learning technique, is one of the data mining technique which can accomplish the data mining task of identifying hidden patterns that exists in students’ performance without having any prior threshold knowledge as inputs. A subset of dimensions from the high dimensional data set with a subset of data points is said to be a subspace. Finding clusters in these subspaces is called subspace clustering. As similar to most of the real-world data sets which are high dimensional, the student data set considered for the study, having student performance data across different subjects of a course has been also identified to be of high dimensions. Therefore, discovering a quality and meaningful cluster from it becomes a challenging task is identified by authors of pawar (2002), Parsons, et al. (2004), Sembiring et al. (2007) and Woo et al. (2012). The aim of using subspace clustering algorithm in this work is to automatically identify subspaces of the feature space in which clusters exists. The subspace clustering algorithm helps to find good quality clusters in subspaces of high dimensional data without forcing the user to have domain knowledge about the clusters, scalability towards high dimensional data sets, non-presumption of any canonical data distribution, insensitive to the order of input data etc. This study has considered subspace clustering technique to identify hidden patterns, as it can help to analyze the students’ academic performance which includes various competency parameters to assess the interest and understanding of the students to meet the objectives of the course been offered.

Complete Article List

Search this Journal:
Open Access Articles: Forthcoming
Volume 11: 4 Issues (2020): Forthcoming, Available for Pre-Order
Volume 10: 4 Issues (2019)
Volume 9: 4 Issues (2018)
Volume 8: 4 Issues (2017)
Volume 7: 4 Issues (2016)
Volume 6: 4 Issues (2015)
Volume 5: 4 Issues (2014)
Volume 4: 4 Issues (2013)
Volume 3: 4 Issues (2012)
Volume 2: 4 Issues (2011)
Volume 1: 4 Issues (2010)
View Complete Journal Contents Listing