Lala S. Riza, Rabihi Awaludin, Heri Sutarno, Munir, Aji P. Wibawa
Generating sets of examination items in educational assessment is not an easy task to do since it must be ensured that all sets are similar in some aspects, such as characteristics, difficulties, qualities, etc. A trivial way to generate the sets is by shuffling the position of items. However, this manner cannot deeply discover students' capabilities and still provides opportunities for cheating. Therefore, this research is aimed at designing a model for generating sets of items automatically by using a clustering method, namely: fuzzy c-means (FCM). This algorithm is used to build cluster centres from data according to the following parameters: Bloom's taxonomy, difficulties levels, expected response time and others (e.g. story, mathematics or programming questions). After obtaining the cluster centres, members were randomly chosen for each set. To evaluate the proposed model, 636 question items collected from five textbooks of computer networking were used. Then, the results, which are some sets of examination items, were analysed statistically along with discussing from the data-mining perspective to measure the similarity in each set. © 2017 WIETE.
Universitas Pendidikan Indonesia, Bandung, Indonesia; Universitas Negeri Malang, Malang, Indonesia
