Course Material Recommendation System Using Student Learning Behavior and Course Material Complexity Score for Slow Learner Students

Educational data mining (EDM) is not only a process of applying data mining algorithms on academic data. That is a process of exploring and providing solutions at various levels of educational system. That is useful for students, educators, management, and administration for decision-making and preparing futuristic strategies. The proposed EDM framework is motivated to enhance the student academic performance. The focus is on threefold: first to identify the student learning behavior to support the students, and teacher provide the remedial actions on the weak students. We propose clustering method for the study of students’ learning behavior associated with positive and negative outcomes (in exams) by utilizing data mining techniques. Second, there is some classifying approach to characterize student based on performance measure that they earn in examination. Applying supervised classifier on the datasets, we have found significant improvement in results. Finally, the study turns toward developing a technique for recommending appropriate study material by calculating readiness of student and complexity of course material. To achieve the objectives, three models are proposed and combined into one for designing accurate and efficient course material recommendation model. In initial steps, popular data mining algorithms, i.e., K-Means, fuzzy c-means (FCM), and kernel-based FCM (K-FCM), are implemented to cluster students according to their learning behaviors. The comparative study demonstrates that K-FCM-based pattern identification is providing more accuracy with respect to other two algorithms. On the other hand to design a model for student performance prediction, two supervised classifiers, i.e., C4.5 and CART, are implemented. During experiments, we found that the C4.5 decision works well for student performance dataset and for predicting the performance. Using both the components, a course study material recommendation model is proposed. Thus an application is implemented to demonstrate the student’s categorization according to their current performance. That also indicates the learning behavior of the student. On the other hand, the course syllabus is used to offer relevant study material. First the syllabus keywords are extracted. These keywords are used as query to find content from the internet. The identified material is offloaded, and then preprocessing of the contents is carried out. The aim is to recover two kinds of features, first by using the natural language processing (NLP)-based text parser to compute the material’s complexity score. Second feature is calculated using TF-IDF as the content features. The calculated features are working as transaction set for Apriori algorithm. The Apriori algorithm is used for the frequent pattern mining. The obtained contents from this mining help to filter the data according to their content relevance. Secondly, the complexity score of the document is used for suggesting the suitable course study material according to student’s readiness compatibility. The experimental evaluation of the proposed ML-based EDM framework demonstrates effectiveness, for supporting students to getting performance feedback, to enhance their learning, and obtaining relevant and personalized study material. The performance analysis in terms of precision, recall, and F-score demonstrates the accurate outcomes of the recommendation model.