Self-service terminal (SST) management assistance
Two or more simultaneous conditions occurring during two or more Self-Service transactions that require a response from an attendant of a management terminal are evaluated in real time. A suggested or an expected response of the attendant for each transaction is determined based on the evaluation. A priority for each expected response relative to the other expected responses is determined based on the evaluation. A queue of the expected responses in priority order is provided in real time to the management terminal for action on the expected responses in the priority order by the attendant.
1 . A method, comprising:
monitoring video data captured of transaction areas for transactions being processed on self-service terminals (SSTs) using overhead cameras situated above the SSTs;
monitoring transaction data produced by the SSTs during the transactions;
evaluating, using a first machine-learning algorithm, the video data and the transaction data for events indicating that an attendant, who is operating a management terminal for the SSTs, is to perform a first action having a first-action priority, the first machine-learning algorithm trained on training video data and training transaction data to detect specific fraud events and to output corresponding first actions and corresponding first-action priorities, wherein the first machine-learning algorithm configures factors and weights to derive a model that produces a confidence value indicative of whether a given transaction is associated with fraud;
evaluating, using a second machine-learning model, the transaction data for transaction interrupts raised by the SSTs indicating that the attendant is to perform a second action having a second-action priority, wherein the second machine-learning algorithm is trained on available transaction data and available transaction interrupts to output corresponding second actions and corresponding second-action priorities;
identifying particular transactions from the transactions that are associated with one or more of the first action or the second action based on the evaluating of the video data and the transaction data;
generating a queue comprising an entry for each particular transaction along with a corresponding first action or second action, and a corresponding first-action priority, or second-action priority, wherein each entry in the queue is associated with a respective identifier indicative of whether the first action or second action can be performed remotely at the management terminal via a transaction monitor or whether an attendant visit to a corresponding SST is required;
sorting the queue in a priority order based on rules that dynamically adjust priorities considering wait times of transactions;
continuously and in real-time adjusting the queue;
dynamically evaluating a priority of a particular transaction of a customer to highest priority in the queue when a particular wait time of the customer exceeds a maximum wait time set by a retailer;
and providing the queue to the management terminal operated by the attendant.
2 . The method of claim 1 , further comprising:
dynamically resorting the queue in a new priority order based on an additional particular transaction identified as being processed on one of the SSTs.
3 . The method of claim 1 , wherein evaluating the video data further comprises identifying the events based on when conditions detected in the video data indicate one or more of: transaction items associated with greater than a threshold number of items, a particular transaction item associated with an assigned value greater than a threshold value, the particular transaction item not being accounted for in corresponding transaction data, a customer associated with a particular transaction being detected as not having paid for the particular transaction and appearing to be preparing to leave a corresponding transaction area.
4 . The method of claim 1 , wherein evaluating the video data further comprises processing the first machine-learning algorithm with the video data and the transaction data as input and receiving the events, the first action, and the first-action priority as output from the first machine-learning algorithm.
5 . The method of claim 4 , wherein evaluating the transaction data and the transaction interrupts further comprises providing as input to the second machine-learning algorithm the transaction data and the transaction interrupts and receiving the second action and the second-action priority as output from the second machine-learning algorithm.
6 . The method of claim 5 , wherein identifying further comprises adjusting an assigned priority of each particular transaction based on the rules in view of remaining particular transactions and corresponding assigned priorities.