Social Network Metrics Integration into Fuzzy Expert System and Bayesian Network for Better Data Science Solution Performance

Abstract

Basic parameters for social network analysis comprise social network common metrics. There are numerous social network metrics. During the data analysis stage, the analyst combines different metrics to search for interesting patterns. This process can be exhaustive with regard to the numerous potential combinations and how we can combine different metrics. In addition, other, non-network measures can be observed together with social network metrics. This chapter illustrates the proposed methodology for fraud detection systems in the insurance industry, where the fuzzy expert system and the Bayesian network was the basis for an analytical platform, and social network metrics were used as part of the solution to improve performance. The solution developed shows the importance of integrated social network metrics as a contribution towards better accuracy in fraud detection. This chapter describes a case study with a description of the phases of the process, from data preparation, attribute selection, model development to predictive power evaluation. As a result, from the empirical result, it is evident that the use of social network metrics within Bayesian networks and fuzzy expert systems significantly increases the predictive power of the model.