An Evolutionary Computation Approach for Automatically Constructing Concept Map to Evaluate the Quality of Test Question Establishment

Ta-Cheng Chen, National Formosa University
Peng-Sheng You, National Chiayi University
Yi-Chih Hsieh, National Formosa University

ABSTRACT
Learning has occurred and where invalid or the wrong concepts are held by the students. A good concept map is very important to describe students' learning portfolio. However, most learning concept maps have to be formed through the suggestions of experts or scholars in their related fields. The concept map can also be built through the test analysis process and the relationship between learners and concepts. But the issues of how to find the optimal association rule for building the concept map are not considered in the past studies. In this study, a novel data mining based approach has been proposed for constructing a most appropriate concept map in the study. Moreover, the logic of contraposition is applied in the proposed approach to optimize the fuzzy membership function so as to dig out the optimal fuzzy association rules. Based on the proposed approach, it is to investigate that whether the test question banks provided by the elementary school textbook publishers are appropriate or not. In other words, it is to know whether the test question establishment based on the test bank is suitable to evaluate studentsˇ¦ achievement or not. According to the proposed approach, it is able to find out the important association rules to form the learning concept map. The test questions can be then evaluated to see whether the confidence, difficulty and discrimination are appropriately considered in the test questions. The experimental result demonstrates that the well organized test problems are with better concept map similarity to the ideal concept map.

(Return to Program Resources)

Updated 07/09/2013