An Implementation of K-NN Classification Algorithm for Detecting Impersonators in Online Examination Environment

Abstract

The online examination platforms also known as computer-based testing (CBT) platforms for conducting mass-driven examinations over computer networks to eliminate certain issues such as delay in marking, misplacement of scripts, monitoring, etc., associated with the conventional Pen and Paper Type (PPT) of examination have also been bedeviled with the issue of impersonation commonly associated with the PPT system. The existing online examination platforms rely on passive mechanisms such as the CCTV system and the human invigilators for monitoring the examination halls against cheating and impersonation. The proposed model integrates some level of intelligence into existing online examination prototype by designing and developing an intelligent agent service that could assess students against impersonation threat in an online examination environment using the K-Nearest Neighbor (K-NN) machine learning classification technique considering the level of accuracy and response time in answering the questions. A total of 3,083 dataset was downloaded from an online repository; 80% (2,466) of the dataset was used for training the model, while 20% (617) dataset was used in testing the model to enable the model detect unseen data correctly. Results showed that the developed model has a 99.99% accuracy rate, precision, recall and f-score.

Adaptive Learner-CBT with Secured Fault-Tolerant and Resumption Capability for Nigerian Universities

“…Automated resumption of tests will facilitate the adoption of Web based examinations by institutions that do not require uninterrupted power or Internet connection. This agrees with [80]- [82].…”

Untitled

“…Examination malpractices are common everywhere and every examination season witnesses the emergence of new and ingenious ways of cheating [7]. According to [8] existing examination systems are mainly concerned with image analysis techniques and biometric systems for identification, recognition, and classification of the candidates which might generally be prone to errors and easily bypassed [9], [10].…”