Large decision intelligence model system and method
A decision support system and a computer-implemented method of enterprise decision support include a data consolidation module, domain-specific machine learning models, an enterprise decision intelligence model, and a user interface layer. The data consolidation module collects data from both internal and external data sources. The domain-specific machine learning models generate decisions based on the collected data. The enterprise decision intelligence model integrates real-time trends and the decisions to provide context-aware recommendations. The enterprise decision intelligence model maintains a decision graph that connects decision variables of the domain-specific machine learning models in a causal relationship, with the domain-specific machine learning models interacting as an interconnected network. The decisions are influenced by affects of decision variables from other domain-specific machine learning models. The user interface layer facilitates interactive decision-making by way of the enterprise decision intelligence model and visualizing the recommendations.
1. A decision support system for an enterprise, comprising:
server processing circuitry configured with
a data consolidation module to collect data from a plurality of internal data sources and at least one external data source that is external to the enterprise;
a plurality of domain-specific machine learning models to generate respective domain-specific outputs using the collected data;
an enterprise decision intelligence model that dynamically incorporates real-time trends and outputs from the plurality of domain-specific machine learning models, and leveraging the outputs from at least two of the domain-specific machine learning models to generate real-time, context-aware recommendations,
wherein the enterprise decision intelligence model utilizes a decision graph to establish causal relationships between decision variables of the domain-specific machine learning models, enabling the domain-specific machine learning models to operate as an interconnected network guided by the decision graph,
wherein the domain-specific machine learning models generate the respective domain-specific decisions based on effects of the decision variables from other domain-specific machine learning models;
a computing device having a user interface layer that facilitates interactive decision-making through the enterprise decision intelligence model and enables visualization of the recommendations, wherein the user interface layer includes an interface for entering overrides for operation of the enterprise decision intelligence model to override a decision by a selected one of the plurality of domain-specific machine learning models; and
a network connecting the server processing circuitry and the computing device so that the interactive-decision making is performed based on the enterprise decision intelligence model of the server processing circuitry,
wherein the enterprise decision intelligence model is configured to perform a utility function that defines criteria that an outcome of a decision is optimized by a decision intelligence module,
wherein the enterprise decision intelligence model is configured to map out decision paths by a decision tree, guided by utility scores determined by the utility function, for generation of optimal recommendations,
wherein the enterprise decision intelligence model includes a self-learning system having a feedback loop that captures outcomes of decisions by the domain-specific machine learning models,
wherein the enterprise decision intelligence model is configured to
track the captured outcomes to determine whether the decisions led to the captured outcomes,
compare the tracked captured outcomes with an expected outcome,
when the enterprise decision intelligence model determines that a captured outcome deviates from the expected outcome, then the enterprise decision intelligence model refines parameters of the corresponding domain-specific machine learning model, and
generate a refined decision using the refined domain-specific machine learning model to make an adjustment to the outcome based on the captured outcome deviated from the expected outcome.
2. The system of claim 1 , wherein the enterprise decision intelligence model includes an algorithm for the utility function that is used to assess an expected utility in terms of enterprise goals, and simulates potential scenarios to optimize decision policies.
3. The system of claim 1 , further comprising an infrastructure layer configured to detect data drift in the collected data and retrain the domain-specific machine learning models.
4. The system of claim 1 , wherein the plurality of domain-specific machine learning models includes a demand forecasting module, an inventory optimization module and a workforce scheduling intelligence module,
wherein the inventory optimization module uses the domain-specific machine learning models to identify optimal inventory levels based on storage costs, lead times, and demand forecasts, and
wherein the demand forecasting module generates a demand decision that effects actions in the inventory optimization module by transmitting forecast insights and necessary inventory adjustments in accordance with the decision graph.
5. The system of claim 1 , wherein the plurality of domain-specific machine learning models includes a demand forecasting module, a marketing intelligence module, and an inventory optimization module,
wherein the domain-specific machine learning models generate the decisions including when the marketing intelligence module generates a decision to increase marketing spending, the enterprise decision intelligence model estimates a causal effect of the decision to increase the marketing spending on demand and inventory levels in the demand forecasting module and the inventory optimization module, respectively.
6. The system of claim 1 , wherein the plurality of domain-specific machine learning models includes a risk management module and an inventory optimization module,
wherein when the risk management module detects a supply chain risk, it alerts the inventory optimization module to adjust inventory levels.
7. The system of claim 1 , wherein the user interface layer includes an interface for entering overrides for operation of the enterprise decision intelligence model to override decisions by specific ones of the plurality of domain-specific machine learning models.
8. The system of claim 1 , wherein the interface for entering overrides includes an interface for modifying or rejecting recommendations generated by the enterprise decision intelligence model before execution of the recommendations in real-time.
9. The system of claim 1 , wherein the self-learning system uses real-time data monitoring tools to track Key Performance Indicators (KPIs) including sales figures, inventory levels, and production efficiency of the decision for capturing the outcomes of the decisions.
10. A computer-implemented method of decision support for an enterprise, comprising:
collecting data from a plurality of internal data sources and at least one external data source that is external to the enterprise;
generating, by a plurality of domain-specific machine learning models, respective domain-specific outputs using the collected data;
dynamically incorporating, by an enterprise decision intelligence model, real-time trends and the outputs from the plurality of domain-specific machine learning models, and leveraging the outputs from at least two of the domain-specific machine learning models to generate real-time, context-aware recommendations,
wherein the enterprise decision intelligence model utilizes a decision graph to establish causal relationships between decision variables of the domain-specific machine learning models, enabling the domain-specific machine learning models to operate as an interconnected network guided by the decision graph,
wherein the domain-specific machine learning models generate the decisions based on effects of the decision variables from other domain-specific machine learning models; and
interacting with a user interface layer that facilitates interactive decision-making through the enterprise decision intelligence model and enables visualization of the recommendations,
entering, by the user interface layer, overrides for operation of the enterprise decision intelligence model to override a decision by a selected one of the plurality of domain-specific machine learning models,
performing, by the enterprise decision intelligence model, a utility function that defines criteria that an outcome of a decision is optimized by a decision intelligence module,
mapping out, by the enterprise decision intelligence model, decision paths by a decision tree, guided by utility scores determined by the utility function, for generation of optimal recommendations,
wherein the enterprise decision intelligence model includes a self-learning system having a feedback loop that captures outcomes of decisions by the domain-specific machine learning models,
further comprising,
tracking the captured outcomes to determine whether the decisions led to the captured outcomes,
comparing the tracked captured outcomes with an expected outcome,
when the enterprise decision intelligence model determines that a captured outcome deviates from the expected outcome, refining parameters of the corresponding domain-specific machine learning model, and
generating a refined decision using the refined domain-specific machine learning model to make an adjustment to the outcome based on the captured outcome deviated from the expected outcome.
11. The method of claim 1 , further comprising
assessing, using an algorithm for the utility function performed by the enterprise decision intelligence model, an expected utility in terms of enterprise goals, and simulating potential scenarios to optimize decision policies.
12. The method of claim 10 , further comprising
detecting, by an infrastructure layer, data drift in the collected data and retraining the domain-specific machine learning models.
13. The method of claim 10 , wherein the plurality of domain-specific machine learning models includes a demand forecasting module, an inventory optimization module and a workforce scheduling intelligence module, the method further comprising:
performing machine learning in the inventory optimization module to identify optimal inventory levels based on storage costs, lead times, and demand forecasts, and
generating, by the demand forecasting module, a demand decision that effects actions in the inventory optimization module by transmitting forecast insights and necessary inventory adjustments in accordance with the decision graph.
14. The method of claim 10 , wherein the plurality of domain-specific machine learning models includes a demand forecasting module, a marketing intelligence module, and an inventory optimization module, the method further comprising
generating, by the domain-specific machine learning models, the decisions including when the marketing intelligence module generates a decision to increase marketing spending, the enterprise decision intelligence model estimates a causal effect of the decision to increase the marketing spending on demand and inventory levels in the demand forecasting module and the inventory optimization module, respectively.
15. The method of claim 10 , wherein the plurality of domain-specific machine learning models includes a risk management module and an inventory optimization module, the method further comprising when the risk management module detects a supply chain risk, it alerts the inventory optimization module to adjust inventory levels.
16. The method of claim 10 , further comprising
entering, by the user interface layer, overrides for operation of the enterprise decision intelligence model to override decisions by specific ones of the plurality of domain-specific machine learning models.
17. The method of claim 10 , further comprising
modifying or rejecting recommendations generated by the enterprise decision intelligence model before execution of the recommendations in real-time.
18. The method of claim 10 , further comprising
tracking, using real-time data monitoring tools, Key Performance Indicators (KPIs) including sales figures, inventory levels, and production efficiency of the decision for capturing the outcomes of the decisions.