Kredītkaršu maksājumu krāpniecisko darījumu atpazīšanas algoritmu salīdzinājums
Loading...
Date
Authors
Advisor
Journal Title
Journal ISSN
Volume Title
Publisher
Latvijas Universitāte
Language
lav
Abstract
Darba mērķis ir salīdzināt krāpniecisko kredītkaršu maksājumu prognozēšanas procesus un veikt praktisko izpēti par tajos izmantotajām metodēm. Darba ietvaros tika izanalizēti plašāk lietotie klasificēšanas algoritmi kredītkaršu datu kontekstā un definēti kvantitatīvi mēri algoritmu salīdzinošai analīzei. Tika veikti praktiski eksperimenti ar darbā aprakstītajām klasifikācijas metodēm un pieejām, analizējot darba autoram pieejamos kredītkaršu maksājumu datus. Detalizēti aprakstīta nepieciešamā datu priekšapstrāde, vēršot uzmanību datu savstarpējai saistībai un nozīmei maksājumu procesos. Tiek analizēta un salīdzināta iegūto modeļu precizitāte, funkcionalitāte un izmantošanas potenciāls. Iegūtie rezultāti parāda katras metodes stiprās un vājās puses, norādot uz novēršamajiem trūkumiem un metožu izmantošanas potenciālu. Darba apjoms - 67 lpp., 6 tabulas, 68 attēli un 7 pielikumi. Atslēgvārdi: kredītkaršu maksājumi, klasifikācijas metodes, pārpratumu matrica, Python programmēšanas valoda
The aim of this master’s thesis is to compare the forecasting processes in case of fraudulent credit card payment detection and to conduct a practical study of the methods used. Most widely used classification algorithms in the context of credit card payment data were analyzed and quantitative measures were defined for the comparative analysis. Practical experiments were conducted with the classification methods and approaches described in the work, analyzing the credit card payment data available to the author. The necessary pre-processing of data was described in detail, focusing on the relationship and relevance of data in payment processes. Analysis, comparison of the accuracy, functionality, and usage potential of the models were conducted. The results obtained show the strengths and weaknesses of each method, indicating the shortcomings to be addressed and the potential for method usage. The work contains 67 p., 6 tables, 68 figures and 7 appendixes. Keywords: credit card payments, classification methods, confusion matrix, Python programming language
The aim of this master’s thesis is to compare the forecasting processes in case of fraudulent credit card payment detection and to conduct a practical study of the methods used. Most widely used classification algorithms in the context of credit card payment data were analyzed and quantitative measures were defined for the comparative analysis. Practical experiments were conducted with the classification methods and approaches described in the work, analyzing the credit card payment data available to the author. The necessary pre-processing of data was described in detail, focusing on the relationship and relevance of data in payment processes. Analysis, comparison of the accuracy, functionality, and usage potential of the models were conducted. The results obtained show the strengths and weaknesses of each method, indicating the shortcomings to be addressed and the potential for method usage. The work contains 67 p., 6 tables, 68 figures and 7 appendixes. Keywords: credit card payments, classification methods, confusion matrix, Python programming language