Digital banking platform and architecture

A computer-implemented method for maintaining electronic representations of aggregate user behavior stored as a plurality of multi-dimensional vectors, each multi-dimensional vector corresponding to a user of a plurality of users and representing an approximation of the user's behavior in n-dimensional space, the method comprising:

• receiving, from one or more point of sale devices, transaction information data sets representing purchase transactions of each user of the plurality of users, the transaction information including, for each purchase transaction, at least a user identifier, an approximate location of purchase, a time-stamp, a retailer, and a price;
• receiving, from one or more harvester applications residing on user computing devices each corresponding to a user of the plurality of users, social media data including at least one of text, video, images, and audio data associated with the transaction information;
• estimating, from the social media data, one or more types of product associated with each purchase transaction;
• generating, for each purchase transaction, a temporary multi-dimensional vector representing the purchase transaction, the temporary multi-dimensional vector storing, in separate dimensions of the vector, numerical values extracted from at least the transaction information and the social data;
• and
• updating, for each user associated with the purchase transactions, an aggregated multi-dimensional vector generated from a corpus of purchase transactions associated for the user by incorporating the temporary multi-dimensional vector into the aggregated multi-dimensional vector;
• wherein the aggregated multi-dimensional vector is adapted for processing by a redemption predictor neural network to identify one or more patterns associated with redemption of one or more electronic offers presented to the plurality of users, the one or more patterns utilized to generate a redemption prediction which is then compared against tracked real-world redemption behavior of the plurality of users to train the redemption predictor neural network.

A non-transitory computer readable media storing a set of computer-interpretable instructions, the set of instructions which, when executed on a processor, cause the processor to perform a computer-implemented method for maintaining electronic representations of aggregate user behavior stored as a plurality of multi-dimensional vectors, each multi-dimensional vector corresponding to a user of a plurality of users and representing an approximation of the user's behavior in n-dimensional space, the method comprising:

• receiving, from one or more point of sale devices, transaction information data sets representing purchase transactions of each user of the plurality of users, the transaction information including, for each purchase transaction, at least a user identifier, an approximate location of purchase, a time-stamp, a retailer, and a price;
• receiving, from one or more harvester applications residing on user computing devices each corresponding to a user of the plurality of users, social media data including at least one of text, video, images, and audio data associated with the transaction information;
• estimating, from the social media data, one or more types of product associated with each purchase transaction;
• generating, for each purchase transaction, a temporary multi-dimensional vector representing the purchase transaction, the temporary multi-dimensional vector storing, in separate dimensions of the vector, numerical values extracted from at least the transaction information and the social data;
• and
• updating, for each user associated with the purchase transactions, an aggregated multi-dimensional vector generated from a corpus of purchase transactions associated for the user by incorporating the temporary multi-dimensional vector into the aggregated multi-dimensional vector;
• wherein the aggregated multi-dimensional vector is adapted for processing by a redemption predictor neural network to identify one or more patterns associated with redemption of one or more electronic offers presented to the plurality of users, the one or more patterns utilized to generate a redemption prediction which is then compared against tracked real-world redemption behavior of the plurality of users to train the redemption predictor neural network.
• wherein each aggregated multi-dimensional vector is represented in an array of a plurality of array elements, and each array element of the plurality of array elements represents a different dimension of the aggregated multi-dimensional vector, the array elements, in combination, representing the approximation of the user's behavior in the n-dimensional space.

A system comprising:

• one or more computer server configured to that receive and transmit one or more data sets over a network;
• one or more client devices with a customer application installed on a mobile device that are connected to the server over the network;
• and
• a computing node connected to one or more blockchains adapted to offload computational requirements of one or more constant block updates, the computing node coupled to one or more financial institutions through a secure connection for transmission of data sets adapted for at least one of banking accounts, peer to peer payments, and payroll management;
• wherein the computing node is configured to determine and parse representations of at least one of user behaviour, information and product information, and to generate one or more comparison vectors and offer recommendations based at least on these representations and transfer the one or more comparison vectors or offer recommendations to a client device of the one or more client devices across the communication network.