Development of Automatic Decision System in the Next Item Selection for Computerized Adaptive Testing

Abstract

The research was invented in order to develop auto-decision system in the next item selection for computerized adaptive testing using Ant colony system with Triangle decision tree (TDT). The process is proposed by three main steps. The first step is dividing group of items from item bank and involved rules design before Ant colony system process. The second step is process design of Ant colony system for make decision
to select the next item properly with ability level of the examinees. The third step is the format operation of Computerized adaptive testing. The efficiency testing of aforementioned steps measures up the examinee ability estimate and True ability, which operate under Monte Carlo Simulation via item bank simulation in accordance with Item Response Theory (IRT). IRT use Three-Parameter Logistic Model (3PL) to simulate the results from item and True ability of the examinees afterwards calculate Root Mean Square Error (RMSE) and Average Bias compare with the next item selection procedure using Maximum Information Criterion (MIC). The results showed that the developed system is more efficient than the next item selection procedure using MIC, the developed system has RMSE = 0.112 and Average Bias = -0.036.