A Novel Fuzzy Logic Model to Identify Closeness for Alias Detection

Abstract

The detection of person alias names are important for improving the accuracy of the information quality. The fuzzy-based decision system is proposed for alias detection, which is a rule-based system that uses fuzzy logic to make a decision about the closeness between the given name pairs. A fuzzy logic is formulated by a set of linguistic variables based on feature’s score value. An entity pair’s association score values are calculated using string and link-based features like Hamming Distance, Leventein Distance, Normalized String Edit Distance, Common Friends, Normalized Dot Product and Co-occurrence Relevance and an output variable accuracy as closeness. These features are transformed into fuzzy input variables and designed with proper membership functions. The proposed novel fuzzy inference system gives the decision of aliases closeness in the form of crisp values ranging from 0 to 1. In this work, the model achieves upto 90% accuracy compared to estimated accuracy.