Identifying Cheating Users in Online Courses
| Autor: | Vicenzo Abichequer Sangalli, Gonzalo Martínez-Muñoz, Estrella Pulido Canabate |
|---|---|
| Rok vydání: | 2020 |
| Předmět: |
Course materials
business.industry Generalization Computer science Cheating 05 social sciences 050301 education Machine learning computer.software_genre Support vector machine 0502 economics and business ComputingMilieux_COMPUTERSANDEDUCATION Artificial intelligence Cluster analysis business 0503 education computer 050203 business & management |
| Zdroj: | EDUCON |
| DOI: | 10.1109/educon45650.2020.9125252 |
| Popis: | Students interact with online courses mainly in two ways: by reviewing the course materials and by solving exercises. However, there are cases in which student behaviour differs and tends to become more focused on solving exercises without looking at course materials. This type of interaction could be an indicative of unethical behavior, such as students who collaborate by sharing answers with one another or fake accounts that are used by students to obtain the correct answers for exercises. In this paper, we propose several metrics to identify these two types of cheating based on co-occurring events and measures of interaction with the course. From the pool of accounts in the course, the pairs of accounts that solve exercises very close in time are considered to be potential collaborating accounts. The proposed metrics are computed for these pairs of accounts and K-means clustering is used to separate pairs of real students who collaborate with respect to students who use fake accounts to harvest the correct answers to exercises. A generalization accuracy over 95% to classify these types of cheating is achieved by using a Support Vector Machine (SVM). |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|