A RISK-BASED SUSPICIOUS FINANCIAL TRANSACTION DETECTION MODEL USING AUTOENCODERS

Abstract

For many years, the financial sector has placed a high priority on identifying questionable financial transactions. Financial companies have historically used rule-based systems to spot possibly fraudulent activity. To identify suspicious activity, these systems use preset criteria and patterns, including big transactions or frequent deposits. Traditional techniques are somewhat successful, but they have several drawbacks. They often produce a high percentage of false positives, necessitating human evaluation of transactions that have been identified. Furthermore, these systems have trouble keeping up with changing fraud trends, which reduces their ability to identify complex financial crimes. Advanced detection systems are now more important than ever due to the increasing number and complexity of financial transactions in the digital age. The dynamic nature of financial fraud is not addressed by traditional solutions, which results in ineffective loss prevention. This makes the need for a more flexible, precise, and scalable method of identifying questionable transactions urgent. Traditional approaches' lack of flexibility and the serious financial and reputational consequences associated with undiscovered fraud highlight the need for a more reliable detection system. Enhancing the capacity to identify unusual patterns in financial data with a low number of false positives while preserving scalability and efficiency is the aim. The suggested system offers a novel approach that makes use of a risk-based evaluation method in conjunction with an autoencoder-based model. By identifying minute deviations from typical patterns in transaction data, this method seeks to identify questionable activity. By including a risk-based framework, the model is guaranteed to take contextual elements into account, minimising false alarms and giving high-risk transactions priority for further examination. By addressing the shortcomings of conventional techniques, this system offers a smart, flexible, and trustworthy instrument for preventing financial fraud.