Despite modernization of financial infrastructure and payment systems, a large share of the Caribbean population remains unbanked, lacking access to formal accounts and electronic fund transfer (EFT) systems. This exclusion prevents them from sending or receiving funds electronically, leaving them financially and socially vulnerable and limiting the reach of social safety nets. Governments collect revenue through taxes, duties, fees, and charges, and disburse funds via salaries, pensions, subsidies, and welfare.
For the unbanked, these flows often require physical cheques, which become inaccessible when mobility or logistics are disrupted. Addressing this gap requires innovative solutions such as digital wallets, which securely store payment information and enable transactions without exposing sensitive details. Digital wallets offer convenience but raise concerns about privacy and fraud, deterring adoption. To mitigate these risks, this study proposes anomaly detection techniques to identify unusual transaction patterns that may signal fraud. The primary objective was to design a digital wallet tailored for Caribbean governments, given their central role in managing financial flows and their capacity to drive financial inclusion. A secondary objective was to develop a robust anomaly detection method to enhance security.
To achieve this, the study applied the Univariate Gated Transformer Networks (UGTN) architecture to financial transaction data from a digital wallet, successfully predicting anomalies. Based on these results, the study recommends integrating the UGTN model into the backend of the proposed wallet to strengthen fraud detection and protect users’ funds. Such integration would enhance trust in mobile payment systems, expand access to EFT, and reduce vulnerability among unbanked populations. In doing so, the study contributes to regional knowledge on digital wallets and financial inclusion, offering a pathway to greater resilience in the Caribbean’s financial ecosystem.
