Credit evaluation using eigenface method for mobile telephone customers

Abstract

Most mobile communication companies in China were puzzled by the arrearage problem. Then credit evaluation for the mobile telephone customer becomes increasingly important in China's telecommunications services industry on economic operation and information management. The study presents a customer credit measure method based on customer attributes by employing the eigenface technology. Analysis of customer credit evaluation indicates potentiality of the proposed eigenface-based credit evaluation method that it provides a potential viable solution to the arrearage problem. By means of principal component analysis, customers represented by their attribute vectors in training set are used to build the eigenspace. Subspace of the eigenspace for credit evaluation is generated by selecting principal orthogonal eigenvectors of covariance matrix of the training samples. Each one of the principal eigenvectors ever called “eigenfaces” is taken as one basis reference customer (component) for credit evaluation. Then each customer represented its feature vector is projected onto the subspace and described by a linear combination of the basis reference customers represented by the selected principal eigenvectors in the subspace. Weights of the linear combination are denoted as one weight vector. Difference between the weight vectors is used to evaluate customer credit. The proposed method yields satisfying results in its application to credit evaluation for 400,000 customers in a mobile communication services company in China. Result analysis indicates that further arrearage management based on credit evaluation is workable.