ahp algorithm and un-supervised clustering in auto insurance fraud detection

this thesis is a study on insurance fraud in iran automobile insurance industry and explores the usage of expert linkage between un-supervised clustering and analytical hierarchy process(ahp), and renders the findings from applying these algorithms for automobile insurance claim fraud detection. the expert linkage determination objective function plan provides us with a way to determine which claims are fraudulent in this model and which input is outlier like fraud. empirical assessment is based on a data set of auto claims occurred in iran – alborz insurance company during 2008-2009. based on indices and anomaly analysis some outliers are fraud in insurance market. the main hypothesis of this research is existence of a relationship between the ahp, clustering and outlier analysis algorithms in their results and an expert system can indicate the truth of this hypothesis. this research concludes that there is a significant relationship between these used methods. ahp can be a parallel and efficient algorithm for solving the statistical problems same as data mining method. the time of occurrence of claim (interval between the time of occurred accident and policy start date) has highest level of important in fraud detection and the insured without any insurance record has a higher level of risk in comparison with the insured that has one or two previous claims also the probability of fraud and to be in subsection groups goes up if the premium of insured increases.