A Proposed Technique for Cheating Detection in MCQ Test based on the K-means Method in an Adaptive E-learning System

The most important problem facing adaptive e-learning platforms is cheating in exams and the difficulty of detecting cheating cases and making recommendations for these cheating cases. In this research paper, a technique to detect cheating in MCQ tests is proposed in a proposed adaptive e-learning system, where web usage mining techniques and the k-means algorithm with Levenshtein distance are used to detect cheating by dividing the learners into clusters according to the similarity in the numbers of their choices in the MCQ test. Also using the Levenshtein distance to make a comparison within each cluster between the number series chosen by the learners to show the corresponding learners. The IQ ratio among the apparent learners from the matching process is used to make recommendations for cheating cases. The data set used to test the proposed technique is two data sets. The first data set is a proposed data set and the second data set is a real data set for learners taking MCQ exams in the Department of Computer Science at the University of Technology. When testing the efficiency of this proposed technique, the measures of performance as it comes: Accuracy is 98.182 %, Precision is 100 %, Recall is 98.182 % and F1-measure is 99.1 %.