A critical assessment of imbalanced class distribution problem: The case of predicting freshmen student attrition

Thammasiri, Dech; Delen, Dursun; Meesad, Phayung; Kasap, Nihat

doi:10.1016/j.eswa.2013.07.046

Cited by 157 publications

(91 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Alkhasawneh, (2011) utilizes neural networks to predict first year retention and provides an extensive analysis of his models' performance. Finally, we highlight the recent work by Thammasiri et al (2013), in which the problem of predicting freshmen student attrition is approached from a class imbalance perspective, and the authors show how oversampling methods can enhance prediction accuracy.…”

Section: Related Workmentioning

confidence: 99%

“…This can essentially decrease the rate of imbalance and it often lessens the bias towards the majority class. Previous work by Thammasiri et al, (2013) has investigated the effect of various over-sampling techniques on models designed to predict student attrition. One such method that was found to improve the overall performance was SMOTE (Chawla, Bowyer, Hall, & Kegelmeyer, 2002), which works by generating synthetic instances by interpolating existing minority class occurrences with their k-nearest neighbours of the same class.…”

Section: Improving the Detection Of At-risk Students With Data Oversamentioning

confidence: 99%

“…It is worth noting that this particular dataset has a highly imbalanced class distribution wherein only 11.5 percent of the instances belong to the minority class (student not retained). As described in Thammasiri et al (2013), predicting student retention becomes more challenging when the available training sets are imbalanced because standard classification algorithms usually have a bias towards the majority class.…”

Section: Descriptionmentioning

confidence: 99%

“…Consequently, a clear focus has been directed towards early identification and diagnosis of at-risk students, and a variety of studies using statistical methods, data ISSN 1929-7750 (online). The Journal of Learning Analytics works under a Creative Commons License, Attribution -NonCommercial-NoDerivs 3.0 Unported (CC BY-NC-ND 3.0) mining, and machine learning techniques can be found in recent literature (Thammasiri, Delen, Meesad, & Kasap, 2013;DeBerard, Spielmans, & Julka, 2004;Zhang, Anderson, Ohland, & Thorndyke, 2004;Burtner, 2005;Yu, DiGangi, Jannasch-Pennell, Lo, & Kaprolet, 2007;Mendez, Buskirk, Lohr, & Haag, 2008;Li, Swaminathan, & Tang, 2009;Lin, Imbrie, & Reid, 2009;Delen, 2010;Zhang, Oussena, Clark, & Hyensook, 2010).…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Engagement vs Performance: Using Electronic Portfolios to Predict First Semester Engineering Student Persistence

et al. 2014

View full text Add to dashboard Cite

As providers of higher education begin to harness the power of big data analytics, one very fitting application for these new techniques is that of predicting student attrition. The ability to pinpoint students who might soon decide to drop out, or who may be following a suboptimal path to success, allows those in charge not only to understand the causes for this undesired outcome, but provides room for the development of early intervention systems. While making such inferences based on academic performance data alone is certainly possible, we claim that in many cases there is no substantial correlation between how well a student performs and his/her decision to withdraw. This is especially true when the overall set of students has a relatively similar academic performance. To address this issue, we derive measurements of engagement from students' electronic portfolios and show how these features can be used effectively to augment the quality of predictions.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Improving the Detection Of At-risk Students With Data Oversamentioning

confidence: 99%

Section: Descriptionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Engagement vs Performance: Using Electronic Portfolios to Predict First Semester Engineering Student Persistence

et al. 2014

View full text Add to dashboard Cite

show abstract

“…However, the imbalanced data can distort the real performance of the prediction model. In other words, a model based on imbalanced data can yield high overall predictive accuracy driven by the majority class but with poor accuracy for the minority class [20]. Therefore, this study employed the over-sampling method SMOTE to handle this problem.…”

Section: Pre-processing For Dealing With the Imbalanced Data Problemmentioning

confidence: 99%

Comparison of Single Classifier Models for Predicting Long-term Business Failure of Construction Companies Using Finance-based Definition of the Failure

Choi¹,

Son²,

Kim³

et al. 2017

Proceedings of the International Symposium on Automation and Robotics in Construction (IAARC)

View full text Add to dashboard Cite

This study compares the performance of six classifier models (ANN, KNN, C4.5, SVM, LR, and NB) for predicting the business failure of construction companies after three years from 2010 to 2012. Although previous studies have explored numerous business failure prediction models for construction companies, these models have focused on short-term failure, defined as failure occurring within one year, and have defined business failure based on companies' experiencing serious legal events, including bankruptcy, delisting, and default. However, the construction industry is typically characterized by projects with longer durations, usually exceeding one year. This implies that previous short-term models cannot predict the business failure of construction companies until the end of particular projects. Moreover, this problem is compounded by the fact that legal events can involve lengthy proceedings, which are often initiated much later than the actual moment of the business failure. Therefore, in this study, six classifier models will be used to predict the business failure of the construction companies within three years using a finance-based definition of failure. The results show that all six models' performances noticeably decrease when they predict more than one year. These results demonstrate that previous short-term prediction models with outstanding performance cannot be practical in predicting the long-term business failure of construction companies.

show abstract

Imbalanced Data Classification Using a Relevant Information-Based Sampling Approach

Hoyos-Osorio

Fernández

Martínez

et al. 2018

Progress in Artificial Intelligence and Pattern Recognition

View full text Add to dashboard Cite

A critical assessment of imbalanced class distribution problem: The case of predicting freshmen student attrition

Cited by 157 publications

References 33 publications

Engagement vs Performance: Using Electronic Portfolios to Predict First Semester Engineering Student Persistence

Engagement vs Performance: Using Electronic Portfolios to Predict First Semester Engineering Student Persistence

Comparison of Single Classifier Models for Predicting Long-term Business Failure of Construction Companies Using Finance-based Definition of the Failure

Imbalanced Data Classification Using a Relevant Information-Based Sampling Approach

Contact Info

Product

Resources

About