Data-driven machine-learning theft detection
A machine-learning algorithm is trained with features relevant to basket data for items of transactions. The trained algorithm is trained to predict whether a given transaction is more or less likely to be associated with theft being engaged in by a transaction operator for the transaction. The trained algorithm is then provided basket data for a given transaction and produces as output a theft prediction value. When the theft prediction value exceeds a configured threshold value, the transaction is flagged for manual intervention or the transaction is flagged for subsequent manual verification.
1. A method, comprising:
providing a processor executable instructions from a non-transitory computer-readable medium causing the processor to perform operations, comprising:
training a machine-learning algorithm on basket data for items of transactions, each basket data includes features that comprise item identifiers for the corresponding items, item classifications for the corresponding items, cluster identifications for the corresponding item classifications, a total number of the corresponding items, a total number of voids for the corresponding transaction, scan interval between recording a previous item and a next item in the corresponding transaction, a terminal type of the transaction terminal associated with the corresponding transaction, a type of transaction for the corresponding transaction, an operator identifier for an operator of the corresponding transaction terminal for the corresponding transaction, average voided item price for any voided items in the corresponding transaction, a total number of items having a quantity greater than 1 in the corresponding transaction, transaction duration for the corresponding transaction, and item price variance for the corresponding items recorded as a variance between a catalogue price of a given item in the corresponding transaction and a recorded price for the given item during the corresponding transaction, wherein training further includes flagging each transaction with an indication as to whether the corresponding transaction was or was not associated with a sweet hearting theft, wherein training further includes training the machine-learning algorithm to use the features as input and produce a sweet hearing prediction value as a value between 0 and 1 where 0 indicated there is no sweet hearting theft and 1 is sweet hearting theft;
producing a trained machine-learning algorithm from the training;
providing the trained machine-learning algorithm with given features derived from a given-set of basket data for a given transaction;
receiving from the trained machine-learning algorithm a given sweet hearting prediction value indicating whether a cashier associated with the given transaction is engaged in the sweet hearting theft; and
integrating the given sweet hearting value into a transaction workflow that processes the given transaction, the transaction workflow performs exception processing when the given sweet hearting value is greater than a threshold value.
2. The method of claim 1 further comprising, retraining the trained machine-learning algorithm with an additional indication as to whether the given transaction was associated with actual theft by the cashier or no theft by the cashier.
3. The method of claim 1 further comprising, linking a flagged transaction that was flagged for the given transaction as being associated the sweet hearting theft to corresponding video associated with the given transaction, wherein the video was captured for the transaction.
4. The method of claim 1 further comprising, interrupting, by the exception processing, the given transaction and preventing the given transaction from concluding until an override is received after an audit of the given transaction.
5. The method of claim 1 further comprising, generating a report associated with the cashier that includes the given transaction and flagged transactions for the cashier when a total number of flagged transactions exceeds a configured amount within a given period of time.