An Item Selection Strategy Based on Association Rules and Genetic Algorithms

The main purpose of academic testing is to improve learning. The computer-based test (CBT) and online test (OLT) have been important trends in e-learning. Many online test systems randomly generate test papers from an item bank. A high-quality test must to consider the following questions. Is the depth and breadth of test items appropriate? Can test items examine student ability at different cogitative levels? Can test items avoid relationships among test items? Can a test identify student ability and provide learning suggestions appropriate? Therefore, it is the important issue to solve above problems by using information technology. This study applies a novel item selection strategy implemented by computer and is based on assessment theory, data mining, genetic algorithms and a revised bloom taxonomy. The proposed strategy ensures that tests are of high quality.