| CPC G06Q 40/06 (2013.01) [G06N 5/025 (2013.01); G06Q 40/03 (2023.01)] | 19 Claims |
|
1. A system for generating and transmitting alert notifications to devices for decision making support based on holistic representations of client finances, the system comprising:
a memory storing a machine learning model structure and state space, the state space defined by variables measuring population goals and user behaviour, the machine learning model structure being a parameterized, normalized, and rational client-centric model, the machine learning model structure having model parameters;
a processor coupled to the memory programmed with executable instructions, the instructions configuring the processor to:
generate event rules conditional on client profile data and prediction data using the machine learning model structure,
detect an event using the event rules and streaming population data, and
an alert engine being configured by machine learning to compute predictive next best action data for a client based on the detected event and model state, the next best action data having evidence supporting the detected event; wherein the processor causes the alert engine to compute the next best action data using supervised learning to predict if a past state of the state space is indicative of a future event based on indicator variables;
generate the client profile data and the prediction data using categorized, normalized, enriched or enhanced features extracted from the streaming population data and to generate a data structure concerning a financial state of the client;
generate visualized data of the next best action data and the financial state of the client;
provide the visualized data to be displayed on the user interface application as an alert notification, and
collect feedback data through a user interface application concerning the next best action data, the feedback data linked to an operational context, and adjust the machine learning model structure and constraints on the model parameters, wherein the processor implements machine learning from the feedback data to tune notification thresholds to improve relevancy of future alert messages, wherein the user interface application is configured to receive the feedback data for tuning and machine learning, wherein the next best action data is predictive and the feedback data is indicative of a relevance variable for the next best action data, the feedback data comprising data indicating the relevance or utility of the next best action data;
wherein the processor updates parameters of the machine learning model that generates the event rules and next best action data based on the feedback data, wherein the feedback data is used to adjust a rank or priority of the next best action data; and
wherein the processor generates a binary univariate thresholding function for features and configures the event rules using the binary univariate thresholding function, wherein the binary univariate thresholding function for features determines a level of saliency or significance in triggering the event detection and the next best action data.
|