Efficient parallelized computation of global behavior profiles in real-time transaction scoring systems

A system comprising a computer program product that includes a non-transitory machine-readable medium storing instructions that, when executed by at least one programmable processor, cause the at least one programmable processor to perform operations comprising:

• dividing a global profile computing component of a transaction scoring model of a real-time behavioral analytics transaction scoring system into a plurality of global profile component instances, the transaction scoring model using a plurality of global profile variables, at least one of the plurality of global profile component instances using its own global profile partition that contains the estimate of global profile variables and being configured for update by a dedicated thread of execution, at least one dedicated thread being configured for receiving and scoring a portion of input transactions;
• partitioning the input transactions for receipt across the plurality of global profile component instances;
• and
• updating at least one of the plurality of global profile partitions by a corresponding global profile component instance running in the dedicated thread according to the scoring algorithm;
• merging the plurality of global profile partitions and individual estimates of global profile variables, without applying a global lock across the plurality of global profile partitions, based on an upweighting of observations in the plurality of global profile partitions by at least one of the plurality of global profile component instances, the upweighting comprising adjusting weights associated with one or more observations by increasing the weight of an observation according to at least one of a number of events observed since a previously performed merge, and an amount of time elapsed since the previously performed merge.

A computer-implemented system for parallelized processing of data by a real-time transaction scoring system to reduce level of contention associated with parallel processing of a plurality of behavior profiles for transacting entities, the system comprising one or more processors configured for:

• dividing a global profile computing component of a transaction scoring model of a real-time transaction scoring system into a plurality of global profile component instances, the transaction scoring model using a plurality of global profile variables, one or more of the plurality of global profile component instances using at least one of a plurality of global profile partitions that contains the estimate of global profile variables and being configured for update by a dedicated thread of execution of the real-time transaction scoring system, the dedicated thread of execution being configured for receiving and scoring a portion of input transactions;
• partitioning, based on one or more transaction routing shuffling algorithms, the input transactions for receipt across the plurality of global profile component instances;
• updating one or more of the plurality of global profile partitions by the corresponding global profile component instances running in the dedicated thread according to the scoring algorithm, the updating being triggered independently from other component instances without a global locking or synchronization, and merging the plurality of global profile partitions and individual estimates of global profile variables, without applying a global lock across the plurality of global profile partitions, based on an upweighting of observations in the plurality of global profile partitions by at least one of the plurality of global profile component instances, the upweighting comprising adjusting weights associated with one or more observations by increasing the weight of an observation according to at least one of a number of events observed since a previously performed merge, and an amount of time elapsed since the previously performed merge, the merging triggered by applying different suitable criteria, and the global profile variables including global profile variable types of one or more of an event rate, daily rate, quantile estimators, and/or behavior sorted lists.