Developing fraud detection models always has been challenging area. Low frequency of fraudulent cases within data, indications instead of certainty contribute to very challenging area for data science method applying. Traditional approach of predictive modelling became insufficient, because relaying on few variables as a base of the fraud model are very fragile concept. Reason for that is fact that we are talking about portfolio with low cases of events, and from the other hand it is unrealistic to lean on few variables articulated through logistic regression, neural network or similar method that will be able to detect sophisticated try of fraudulent activities. Chapter gives proposal how to use data science in such situations where there are no solid bases but only potential suspicious regarding fraudulent activities. For those purposes Benford’s law in combination with other data science methods and fuzzy logic will be used on sample data set, and will be shown potentials of proposed methodology for fraud detection purposes. Chapter shows case study in domain of finance on public data, where proposed methodology will be illustrated an efficient methodology which can be usable for fraud detection purposes.
CITATION STYLE
Klepac, G. (2018). Cognitive data science automatic fraud detection solution, based on benford’s law, fuzzy logic with elements of machine learning. In Lecture Notes on Data Engineering and Communications Technologies (Vol. 14, pp. 79–95). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-319-70688-7_4
Mendeley helps you to discover research relevant for your work.