Abstract: Embodiments provide methods and systems for reducing bias in an artificial intelligence model. A method includes computing, by a processor, a reward value based at least in part on a similarity between model predictions from a pre-trained model and agent predictions from a Reinforcement Learning (RL) agent. The method includes performing each step of one or more steps of a rule of a plurality of rules. The rule is assigned a weight and the rule includes a protected attribute, a cumulative statistic value type, and a comparison threshold. The method includes sending a cumulative reward value generated using the reward value and each weighted punishment value computed based at least in part on applying each rule of the plurality of rules to the RL agent. The RL agent learns to biases from the agent predictions while maintaining similarity with model predictions by maximizing the cumulative reward value.

Abstract: Systems and computer-implemented methods are described for modeling card inactivity. For example, hierarchical modeling may be used in which a first level classifier may be trained and validated to predict whether a card will be inactive. For cards predicted to become inactive by the first level classifier, a second level classifier may be trained and validated to predict when the card will become inactive. The first level classifier may include a binary classifier that generates two probabilities that respectively predict that the card will and will not become inactive. The second level classifier may include a multi-class classifier that generates a first probability that the card will become inactive at a first time period (such as one or more months in the future) and a second probability that the card will become inactive at a second time period. The multi-class classifier may generate other probabilities corresponding to other time periods.

Abstract: A growth predictor includes a monitor, a prediction engine, and a prioritization engine. The monitor receives or generates first information of a network already identified as a candidate money laundering (ML) network by an anti-money-laundering system. The prediction engine predicts second information indicative of a growth size of the ML network at a future time based on the first information. The prediction engine executes one or more predictive models to generate the second information indicative of growth size based on the first information, which indicates one or more changes that have occurred in the candidate ML network over a past period of time. The prioritization engine determines a priority of the candidate ML network based on the second information.

Abstract: Embodiments provide methods and systems for predicting chargeback behavioral data of an account holder. The method performed by a server system includes accessing payment transaction data associated with the account holder from a transaction database. The payment transaction data includes a set of transaction indicators corresponding to payment transactions performed by the account holder within a predetermined time period. The method further includes generating a set of transaction features based on the set of transaction indicators. Furthermore, the method includes computing, via a chargeback risk prediction model, a set of chargeback risk probability scores corresponding to one or more time intervals associated with the account holder based, at least in part, on the set of transaction features. The method also includes transmitting a notification to an issuer server associated with the account holder based, at least in part, on the set of chargeback risk probability scores.

Abstract: Systems and computer-implemented methods are described for modeling card inactivity. For example, hierarchical modeling may be used in which a first level classifier may be trained and validated to predict whether a card will be inactive. For cards predicted to become inactive by the first level classifier, a second level classifier may be trained and validated to predict when the card will become inactive. The first level classifier may include a binary classifier that generates two probabilities that respectively predict that the card will and will not become inactive. The second level classifier may include a multi-class classifier that generates a first probability that the card will become inactive at a first time period (such as one or more months in the future) and a second probability that the card will become inactive at a second time period. The multi-class classifier may generate other probabilities corresponding to other time periods.

Abstract: Embodiments provide methods and systems for detecting frauds in payment transactions made by payment instrument using spend patterns of multiple payment instruments associated with user. The method performed by server system includes accessing payment transaction data associated with a plurality of customers from a transaction database. The method includes training a first generative adversarial network (GAN) model based on the payment transaction data and a plurality of probable fraud risk conditions. The first GAN model is trained to generate simulated customer fraud behaviors. The method includes filtering, by the server system, the simulated customer fraud behaviors based on a predetermined filtering criteria. The method includes generating, by the server system, fraud risk scores for the simulated customer fraud behaviors based on a fraud risk model. The method includes extracting fraud risk rules based on a set of simulated customer fraud behaviors from the simulated customer fraud behaviors.

Abstract: Embodiments provide methods and systems for reducing bias in an artificial intelligence model. A method includes computing, by a processor, a reward value based at least in part on a similarity between model predictions from a pre-trained model and agent predictions from a Reinforcement Learning (RL) agent. The method includes performing each step of one or more steps of a rule of a plurality of rules. The rule is assigned a weight and the rule includes a protected attribute, a cumulative statistic value type, and a comparison threshold. The method includes sending a cumulative reward value generated using the reward value and each weighted punishment value computed based at least in part on applying each rule of the plurality of rules to the RL agent. The RL agent learns to biases from the agent predictions while maintaining similarity with model predictions by maximizing the cumulative reward value.