Organized retail crimes and fraud detection for e-commerce marketplace platforms

The research field of this thesis focuses on organized retail crime (ORC) and its impact on e-commerce marketplace platforms, specifically in the context of fraud detection. Despite the increasing prevalence of ORC, there remains a gap in the literature regarding effective fraud detection mechanisms tailored to the unique challenges of online retail environments. This thesis aims to address this gap by investigating and developing robust fraud detection methods for e-commerce platforms. The objectives of the thesis are threefold: first, to conduct a systematic literature review (SLR) to identify gaps and research questions in the field; second, to develop numeric-based machine learning (ML) models for fraud detection, with a focus on handling imbalanced datasets; and third, to explore novel methodologies for text-based fraud detection using imbalanced learning techniques. The methods employed include a comprehensive SLR to synthesize existing research, followed by the development and evaluation of ML models using advanced algorithms and techniques. Additionally, novel methodologies are explored to address the unique challenges of text-based fraud detection, including data augmentation and algorithm development. Key results and findings indicate the efficacy of the proposed ML models in accurately identifying fraudulent activities, particularly in imbalanced datasets. The integration of SHAP analysis enhances model interpretability and transparency, providing valuable insights into feature importance. Furthermore, the text-based fraud detection models demonstrate significant improvements in accuracy and performance. Overall, this thesis contributes to the advancement of fraud detection techniques for e-commerce platforms by addressing critical gaps in the literature and proposing innovative methodologies. By leveraging state-of-the-art ML techniques and enhancing the understanding of organized retail crime, this research aims to bolster the resilience of e-commerce ecosystems against fraudulent activities, thereby fostering trust and confidence among online consumers and businesses.