A Customer Credit Visual Analytics Model Based on Markov Logic Networks

Abstract

Customer credit evaluation is very important for customer relationship management in Enterprise Resource Planning. However, how to evaluate the customers’ credit is a complicated problem. In this paper, we present a Customer Credit Visual Analysis Model (CCVAM) that can be used to evaluate and classify the credibility of new customers according to the historical data about past customers. The model is based on Markov Logic Networks (MLNs) that combines probability and first-order logic with a weight attached to each formula. In this model, the basic rules or indexes based on expert knowledge are transformed into the representation of normal form. Then MLNs is obtained by combining first-order logic and probabilistic graphical models of the rules. After acquiring the weights attached to the rules in a first-order logic knowledge base, the model provides an interface for visualize the corresponding relationships among the rules and weights. The model has been applied to grade clients’ credit degree in an international enterprise and achieved anticipative results. This model can also be used in other areas where level evaluation is required.