Improving the prediction accuracy in blended learning environment using synthetic minority oversampling technique
| DOI | https://doi.org/10.1108/IDD-08-2018-0036 |
| Published date | 20 May 2019 |
| Pages | 76-83 |
| Date | 20 May 2019 |
| Author | Gabrijela Dimic,Dejan Rancic,Nemanja Macek,Petar Spalevic,Vida Drasute |
| Subject Matter | Library & information science,Library & information services,Lending,Document delivery,Collection building & management,Stock revision,Consortia |
Improving the prediction accuracy in blended
learning environment using synthetic minority
oversampling technique
Gabrijela Dimic
School of Electrical and Computer Engineering of Applied Studies, Belgrade, Serbia
Dejan Rancic
Faculty of Electronic Engineering, Niš, Serbia
Nemanja Macek
School of Electrical and Computer Engineering of Applied Studies, Belgrade, Serbia
Petar Spalevic
Singidunum University, Belgrade, Serbia, and
Vida Drasute
Kaunas University of Technology, Kaunas, Lithuania
Abstract
Purpose –This paper aims to deal with the previously unknown prediction accuracy of students’activity pattern in a blended learning environment.
Design/methodology/approach –To extract the most relevant activity feature subset, different feature-selection methods were applied. For
different cardinality subsets, classification models were used in the comparison.
Findings –Experimental evaluation oppose the hypothesis that feature vector dimensionality reduction leads to prediction accuracy
increasing.
Research limitations/implications –Improving prediction accuracy in a described learning environment was based on applying synthetic minority
oversampling technique, which had affected results on correlation-based feature-selection method.
Originality/value –The major contribution of the research is the proposed methodology for selecting the optimal low-cardinal subset of students’
activities and significant prediction accuracy improvement in a blended learning environment.
Keywords Machine learning, Educational data mining, Naïve Bayes classifier, Learning management system, Feature selection,
Oversampling technique
Paper type Research paper
1. Introduction
Over the past years, research studies have applied machine-
learning and data-mining techniques to find interesting
patterns and concepts in educational domain data.
Implementation of data-mining and machine-learning
methodologies in education created a new exploration field
known as educational data mining (EDM) that deals with the
issues in developing methodsfor extrication of knowledge from
data in educational environment(Romero and Ventura, 2007).
EDM process transforms raw data from educational systems
into useful information that could have a great influence on
educational research and practice. EDM uses typical data-
mining techniques of classification, clustering, association
rules, sequential patterns, text mining as well as new
methodologiessuch as discovery of knowledge with models and
integration with psychometrically modeled environment
(Baker and Yacef, 2009).
Previous research studies have shown that performance of
the prediction model highly depends on the selection of
relevant features. As a part of data pre-processing task, feature
selection is defined as a process of identifying and removing
irrelevant and redundant information (Blum and Langley,
1997). In the domain of education, selection of the best
features is of outmost importance, especially in the case of
predictive modeling. Incorrectly disregarded or selected
features may cause complex or unreliable prediction model
generation.Hence, the key problem is how to determine feature
relevance. This can be achieved by applying different feature-
selection techniquesand using classifiers to evaluate the impact
of selected featurestoward the prediction.
The implementation of online learning platforms in
educational environmentshas significantly changed the ways in
The current issue and full text archive of this journal is available on
Emerald Insight at: www.emeraldinsight.com/2398-6247.htm
Information Discovery and Delivery
47/2 (2019) 76–83
© Emerald Publishing Limited [ISSN 2398-6247]
[DOI 10.1108/IDD-08-2018-0036]
Received 27 August 2018
Revised 12 December 2018
Accepted 20 December 2018
76
To continue reading
Request your trial