Model rane detekcije sumnjivih bankovnih transakcija

master's thesis

• Title (english): Model for early detection of suspicious bank transactions
• Author: Darjan Božić
• Mentor: Lucija Rogić Dumančić (mentor)
• Committee member: Jakša Krišto (predsjednik povjerenstva)
• Committee member: Dajana Barbić (član povjerenstva)
• Granter: University of Zagreb; Faculty of Economics and Business; Zagreb
• Defense date and country: 2023-02-21, Croatia
• Scientific / art field, discipline and subdiscipline: SOCIAL SCIENCES; Economics; Finance

Abstract

Volumen oprane količine novca na svjetskoj se razini povećava, zbog čega razvitak novih načina otkrivanja pranja novca postaje prioritet. Ovaj rad nastoji na teorijskoj osnovi razviti model rane detekcije pokušaja pranja novca baziran na skupovima pravila koja procjenjuju sumnjivost bankovnih transakcija te funkcionalnu verziju opisanog modela testirati na generiranom skupu od 10.000 transakcija. Sumnjivim je označeno njih 119, a najčešći razlog za sumnju bili su iznosi iznad određene vrijednosti i transakcije sa rizičnim državama. Dodatno, istraženi su načini poboljšanja modela upotrebom rudarenja podataka i strojnog učenja, koje je u stanju analizirati ogromne količine povijesnih podataka, u njima uočiti uzorke i povezanosti i na temelju njih smjesta otkriti slične transakcije pranja novca, u trenutku njihovog izvršavanja, sa preciznošću do 90%.

Abstract (english)

The volume of laundered money is increasing worldwide, which is why development of new ways of detecting money laundering is becoming a priority. This work aims to, on a theoretical basis, develop a model for the early detection of money laundering attempts based on sets of rules that assess the suspiciousness of bank transactions and to test the functional version of the described model on a generated set of 10,000 transactions. 119 of them were marked as suspicious, and the most common reason for suspicion were amounts above a certain value and transactions with risky countries. Additionally, ways of improving the model were explored using data mining and machine learning, which is able to analyse immense amounts of historical data, spot patterns and connections in them and, based on them, immediately detect similar money laundering transactions, at the time of their execution, with a precision of up to 90 %.

• Keywords: pranje novca; transakcija; otkrivanje; model; strojno učenje
• Keywords (english): money laundering; transaction; detection; model; machine learning
• Language: croatian
• URN:NBN: urn:nbn:hr:148:243751
• Study programme: Title: Business Economics; Study programme type: university; Study level: integrated undergraduate and graduate; Academic / professional title: magistar/magistra ekonomije (magistar/magistra ekonomije)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2023-02-21 14:24:05