Abstract: A computer-implemented method for providing dynamic open banking data aggregation that includes: applying pre-model rules to data regarding financial institution (FI) accounts to determine that a first subset of the FI accounts will be included in an aggregation batch; inputting the data regarding a second subset of the FI accounts to a machine learning (ML) model to determine that a third subset of the FI accounts will be included in the aggregation batch and that a fourth subset of the FI accounts will not be included in the aggregation batch; based on the determination from the application of the pre-model rules and on the determination from the ML model, respectively, requesting data updates for the first and third subsets from the FI; batching the aggregation batch by processing the data updates for the first and third subsets to produce results; and storing the results to an aggregated data store.

Abstract: A system is configured to retrieve a set of raw transaction data. A transaction categorization model is applied to the raw transaction data. The transaction categorization model infers a category from each transaction and labels each transaction with the inferred category. An entity recognition model is applied to the labelled transaction data. The entity recognition model extracts an entity from each transaction and labels each transaction with the extracted entity. The system generates a plurality of transaction streams from the labelled transactions based on the category and entity labels. The system also labels each transaction stream with either a revenue label or a non-revenue label based on an analysis of the types of transactions defining the transaction stream. The system trains a supervised-based neural network using the labelled transaction streams to generate a revenue stream classifier model.

Abstract: A system is configured to retrieve a set of raw transaction data. A transaction categorization model is applied to the raw transaction data. The transaction categorization model infers a category from each transaction and labels each transaction with the inferred category. An entity recognition model is applied to the labelled transaction data. The entity recognition model extracts an entity from each transaction and labels each transaction with the extracted entity. The system generates a plurality of transaction streams from the labelled transactions based on the category and entity labels. The system also labels each transaction stream with either a revenue label or a non-revenue label based on an analysis of the types of transactions defining the transaction stream. The system trains a supervised-based neural network using the labelled transaction streams to generate a revenue stream classifier model.

Abstract: According to disclosed methods and apparatus, a mobile robot moves autonomously along a planned path defined in a coordinate map of a working environment, and dynamically updates the planned path on an ongoing basis, to avoid detected obstacles and projections of detected obstacles. A “projection” arises in the context of moving obstacles detected by the mobile robot, at least in the case for a moving detected obstacle that meets certain minimum requirements, such as minimum speed, persistence, etc. The mobile robot makes a projection by, for example, marking map coordinates or map grid cells as occupied, based not only on the currently detected location of a moving obstacle but further on the most recent estimates of speed and direction. By feeding both detected locations and projections into its path planning algorithm, the mobile robot obtains sophisticated avoidance behavior with respect to moving obstacles.