Archives and Documentation Center
Digital Archives

Network-based methods for anti-money laundering

Show simple item record

dc.contributor Graduate Program in Industrial Engineering.
dc.contributor.advisor Aşıcı, Tınaz Ekim.
dc.contributor.author Taş, Gizem.
dc.date.accessioned 2023-03-16T10:29:55Z
dc.date.available 2023-03-16T10:29:55Z
dc.date.issued 2020.
dc.identifier.other IE 2020 T37
dc.identifier.uri http://digitalarchive.boun.edu.tr/handle/123456789/13431
dc.description.abstract This study concentrates on the contributions of network-based methods in antimoney laundering. Being responsible for reporting suspicious activities, time-consuming analyses conducted by the banks through rule-based systems result in high false positive rates. Today, machine learning methods draw attention for their ability to reduce computational cost and false positive rates in anti-money laundering. In this research, the goal is to both reduce dimension of large and complex financial networks while minimizing the loss of valuable information and identify suspiciousness by feeding machine learning algorithms by network-based features using bank transaction networks. Dimensional reduction of networks in the first place enhanced the efficiency of following subgraph construction algorithms. Based on the hypothesis that money launderers disguise organizedly in licit activities, hence network relations bear the trace of hidden suspects, candidate subgraphs are constructed. Network-based features of each subgraph such as inside transaction volume, amount of outgoing transactions, size and so forth are calculated and these subgraphs are labeled as innocent or suspicious according to presumed risk aversion behaviors of banks. Created candidate subgraphs data set is used for training supervised classifiers and the evaluations on separate test data sets indicate that a satisfactory discriminative performance is obtained. These outcomes support the initial hypothesis that network-based analysis in anti-money laundering would contribute more compared to accustomed rule-based systems or individual analysis methods. These findings are promising for declining false positive rate and easing the burden on labor force in the event of a prospective implementation of the suggested system on real data to be acquired from banks.
dc.format.extent 30 cm.
dc.publisher Thesis (M.S.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2020.
dc.subject.lcsh Money laundering.
dc.subject.lcsh Money laundering -- Prevention.
dc.title Network-based methods for anti-money laundering
dc.format.pages xiii, 96 leaves ;


Files in this item

This item appears in the following Collection(s)

Show simple item record

Search Digital Archive


Browse

My Account