A grouping strategy based partition algorithm for clustering questions in a question bank

Abstract

Grouping algorithms are widely used in multidisciplinary fields such as data mining, image analysis and bioinformatics. This paper proposes the use of Grouping Strategy based Partition Algorithm for clustering the questions in a Question Bank. It includes a new approach for computing the question similarity matrix and use of the matrix in clustering the questions. The grouping algorithm extracts n module-wise questions, compute n × n similarity matrix by performing n × (n-1)/2 pair-wise question vector comparisons and uses the matrix in formulating question clusters. Grouping algorithm has been found efficient in reducing the best case time complexity, O (n× (n-1)/2 log n) of hierarchical approach to O (n × (n-1)/2). Experimental study was carried out by performing a comparative analysis of question clusters formulated with two different similarity measures. Performance evaluation using F-measure proves that grouping strategy based partition algorithm is efficient in formulating question clusters without the initial specification of the number of clusters as well as the iterative stages of cluster formulation.