Application of SVM and Fuzzy Set Theory for Classifying with Incomplete Survey Data

Abstract

Classification with incomplete survey data is a new subject, and also which is an important theme in data mining. This paper proposes a novel, powerful classification machine, support vector machine (SVM) based model of classification for incomplete survey data. Using this model, an incomplete survey data is translated to fuzzy patterns without missing values firstly, and then used these fuzzy patterns as the exemplar set for teaching the support vector machine. Experimental results from the real-world data verify the effectiveness and applicability of the proposed model. Compared with other classification techniques, the method can utilize more information provided by the data, and reveal the risk of the classification result.