Abstract: Fraud detection remains an active area for research. While fraud is a difficult problem hampered by issues ranging from uncertain data to adversarial environments, new technologies and techniques from the fields of data science, machine learning, and big data bring opportunities to alleviate some of the difficulties. In this paper, fraud detection is examined in the context of financial systems for public-sector applications. Public sector financial systems face challenges with fraud detection, such as a large volume of transactions and a need for complex monitoring rules based on the context of transactions. The analysis of public sector financial systems is framed as a data-driven approach to understand the domain. Initial steps to facilitate the analysis of data within a public-sector context are taken by proposing a model framework consisting of detectors, monitors, and pattern mining techniques along with an input data requirements and output results taxonomy. The frameworks’ components are investigated and elaborated on in terms of the public sector financial systems. Furthermore, future directions toward further development and evaluation of the components within the context of public sector financial systems. Fraud detection remains an active area for research. While fraud is a difficult problem hampered by issues ranging from uncertain data to adversarial environments, new technologies and techniques from the fields of data science, machine learning, and big data bring opportunities to alleviate some of the difficulties. In this paper, fraud detection is examined in the context of financial systems for public sector applications. Public sector financial systems face challenges with fraud detection, such as a large volume of transactions and a need for complex monitoring rules based on the context of transactions. The analysis of public sector financial systems is framed as a data-driven approach to understand the domain. Initial steps to facilitate the analysis of data within a public sector context are taken by proposing a model framework consisting of detectors, monitors, and pattern mining techniques, along with an input data requirements and output results taxonomy. The frameworks’ components are investigated and elaborated on in terms of public sector financial systems.
Keywords: Anomaly Detection,Predictive Analytics,Behavioral Modeling,Supervised Learning,Unsupervised Learning,Real-time Monitoring,Data Integration,Risk Scoring,Entity Resolution,Natural Language Processing (NLP),Data Lake Architecture,Feature Engineering,Graph Analytics,Model Explainability (XAI),Regulatory Compliance Analytics.
|
DOI:
10.17148/IJARCCE.2020.91221
