Experimental performance analysis of confidence-based online assessment portal in e-learning using data mining

• Main Authors: Ravikiran, R.K., Anil Kumar, K.R.
• Format: Conference Proceeding
• Language: English
• Subjects: Confidence score; Data mining; e-learning; Machine learning
• ISSN: 2214-7853; 2214-7853
• DOI: 10.1016/j.matpr.2021.04.456

The objective of this research is the creation of a novel confidence-based online assessment portal. Confidence based learning helps us to understand both the accuracy of knowledge in a subject and the amount of confidence in that subject. It helps the learners to improve the confidence level they have in a particular domain or subject. It basically helps to eradicate the simple guessing techniques that learners use in traditional assessment tests. This portal is to be designed such that the learners are made to attend periodic assessment tests. They contain multiple choice questions wherein the learner is made to select one of the four options. In addition, the learner must also select a confidence score which is one among 20, 40, 60, 80 or 100. This score is used to test the level of confidence a learner has while answering an answer. The marks are allotted based on the confidence score. The learner is also awarded negative score based on the wrong answers along with the combination of the confidence score. A dataset is generated that contains the scores obtained by multiple learners. They are then categorized into four grades namely “Below average”, “Average”, “Good” and “Excellent”. The dataset is divided into training, testing and validation groups. Using the training dataset, machine learning tools such as k-nearest neighbour, support vector machine, naïve Bayes, decision tree, etc., are used to create classification models. These models are analysed using the testing dataset and the model that performs the best is selected. Evaluation is done using metrics like precision, recall, F-score, accuracy and specificity. The best model is selected and finally deployed using the validation dataset. This model is used to find the category of a learner. A learner is allowed to obtain a certificate only if he attains the category “Excellent”. Else he is made to relearn the subject and attend the assessment once again until he achieves “Excellent” grade. In this way the e-learning assessment system can be improvised using data mining.