Two classification methods of individuals for educational data and an application

Both methods, Rule Space Method (RSM) and Neural Network Model (NNM), are techniques of statistical pattern recognition and classification approaches developed from different fields - one is for behavioural sciences and the other is for neural sciences. RSM is developed in the domain of educational statistics. It starts from the use of an incidence matrix Q that characterises the underlying cognitive processes and knowledge (Attribute) involved in each Item. It is a grasping method for each examinee's mastered/non-mastered learning level (Knowledge State) from item response patterns. RSM uses multivariate decision theory to classify individuals, and NNM, considered as a nonlinear regression method, uses the middle layer of the network structure as classification results. We have found some similarities and differences between the results from the two approaches, and moreover both methods have characteristics supplemental to each other when applied to the practice. In this paper, we compare both approaches by focusing on the structures of NNM and on knowledge States in RSM. Finally, we show an application result of RSM for a reasoning test in Japan.