Application Of Support Vector Machine To Cognitive Diagnosis

As society develops, people have not satisfied with the achievement that test only gives an evaluation of the macro-level, but hope to get more information about the examinees'cognitive states from the examinees'response in order to make more objective assessment and remedy for individual examinee. Cognitive Diagnostic Assessment could evaluate or diagnose the examinees'cognitive process and psychological process when the examinee finished the test and further more could give remedies. Attribute Hierarchy Model (AHM) proposed by Leighton et.al(2004) is a cognitive diagnostic model. Assuming that items in the test can be characterized by the identified hierarchical ordering of attributes, AHM classifies the observed examinee response patterns to the expected examinee response patterns, and estimates examinees'cognitive states for effective and targeted remedies. Research in this article is carried out under the attribute hierarchy.Support Vector Machine which has a solid theoretical foundation in statistics, has become a closely watched classification technique. SVM use Optimization method to solve the problems of machine learning. Support vector machine can use a small sample to achieve good generalization performance and global optimization. In many practical applications of this technology (such as handwritten digital recognition, text classification, etc.) demonstrated a promising practice .This article applied SVM to the modern educational measurement's diagnostic classification of 0,1 scoring test ,and then compare the classification results with the typical cognitive diagnostic classification .The results show that using SVM to cognitive diagnostic classification, which only need a small sample for training, can ensure a high Pattern match ratio (PMR) and Marginal match ratio(MMR), while required short time to run.The identifying of the items'properties has many problems such as the workload is so heavy that experts are easy to fatigue and the requirements are large. Based on the advantages of SVM shown on the above experiments and Matrix Qt which can be obtained using Matrix R through expansion algorithm exhausts all items, this article propose applying SVM to the identification of items'unknown properties while giving the attributes of critical items which are items in Qt matrix. Experiments' results show high precision and certain feasibility. This has a certain meaning for boosting cognitive diagnostic test.