Sarcouncil Journal of Applied Sciences Aims & Scope

Abstract: Risk profiling using AI has become a topical area in terms of compliance where the volume of transactions, the need to minimize fraud, identity abuse, and increased regulatory pressures are growing. The impossibility to detect the emergent misconduct, hidden abnormalities, and context-specific violation of the compliance rules by fixed rules and regular audits has made the analysis of the behavioural patterns particularly topical. The review forms a conceptual framework of how behavioural risk detection is related to explainable and proportionate compliance remediation. It covers the significant techniques like anomaly detection, sequence learning, graph-based modelling, cost-sensitive classification, and behavioural biometrics and also refers to the problem of explainability, fairness, privacy, and deployment. The literature review has shown that behavioural cues will be effective to increase sensitivity to the emergent threats, especially when dynamics and relational trends are considered. Nevertheless, an imbalanced performance persists in the case of class imbalance, concept drift, delayed labels and institution specific operating conditions. Low interpretability weakens governance structures, low benchmark comparability and weak integration of risk scoring and downstream remediation are some of the weaknesses that have been reported. The advancement in this regard does not simply involve precision in risk identification, but also systems that have the ability to relate the evidence of risks to auditable and visible compliance action that is reasonable.