Mutual information resolution recommendations and graphical visualizations using probabilistic graphical models
A system and method are disclosed for training a probabilistic graphical model based on historical attributes of a supply chain to represent supply chain performance, selecting supply chain entity target variables, collating with the use of machine learning models, a list of features and classes pertaining to selected supply chain entity target variables, calculating first and second level features associated with the list of features and classes, generating supply chain predictions based on the trained probabilistic graphical model, where the supply chain predictions are based on test data, comparing the supply chain predictions to desired supply chain outputs to determine a delta distance, and generating resolution actions, to decrease the delta distance between the supply chain output predictions and the desired supply chain outputs.
1. A computer-implemented method, comprising:
training, by a computer comprising a processor and memory, a probabilistic graphical model based on historical attributes of a supply chain comprising one or more supply chain entities to represent a performance of one or more supply chain entities in the supply chain;
selecting, by the computer, one or more supply chain entity target variables;
collating, by the computer and with a use of one or more machine learning models, a list of one or more features and feature classes pertaining to the one or more selected supply chain entity target variables;
calculating, by the computer, first-level features and second-level features associated with the list of one or more features and feature classes;
generating, by the computer, one or more supply chain output predictions based on the trained probabilistic graphical model, wherein the one or more supply chain output predictions are based on test data;
comparing, by the computer, the one or more supply chain output predictions to one or more desired supply chain outputs to determine a delta distance;
generating, by the computer, one or more resolution actions, based at least in part on the first-level features and second-level features, to decrease the delta distance between the one or more supply chain output predictions and the one or more desired supply chain outputs;
adjusting, by the computer, the trained probabilistic graphical model based, at least in part, on the one or more supply chain output predictions; and
sending, by the computer, instructions to automated machinery, wherein based on the instructions, the automated machinery locates items to add to or remove from an inventory or shipment for the one or more supply chain entities.
2. The computer-implemented method of claim 1 , further comprising:
generating, by the computer, a Bayesian network graph visualization to display a network graph.
3. The computer-implemented method of claim 1 , further comprising:
generating, by the computer, recursive level-ranked features based on the list of one or more features and feature classes; and
displaying, by the computer and on a display device, a network graph comprising the recursive level-ranked features.
4. The computer-implemented method of claim 1 , further comprising:
receiving, by the computer, a query to identify a root cause of a particular metric in the supply chain; and
traversing, by the computer, the probabilistic graphical model to identify features having a largest impact on the particular metric.
5. The computer-implemented method of claim 1 , wherein the one or more supply chain output predictions are based on ensemble data, the ensemble data comprising a combination of the training data and the test data.
6. The computer-implemented method of claim 3 , wherein the ranking for the recursive level-ranked features is based upon a magnitude of change a feature will cause for a KPI or SLA of the supply chain.
7. The computer-implemented method of claim 1 , wherein the historical attributes are normalized, aggregated, or rescaled to allow a direct comparison of data received from different supply chain entities.
8. A system, comprising:
a computer, comprising a processor and memory, the computer configured to:
train a probabilistic graphical model based on historical attributes of a supply chain comprising one or more supply chain entities to represent a performance of one or more supply chain entities in the supply chain;
select one or more supply chain entity target variables;
collate with a use of one or more machine learning models, a list of one or more features and feature classes pertaining to the one or more selected supply chain entity target variables;
calculate first-level features and second-level features associated with the list of one or more features and feature classes;
generate one or more supply chain output predictions based on the trained probabilistic graphical model, wherein the one or more supply chain output predictions are based on test data;
compare the one or more supply chain output predictions to one or more desired supply chain outputs to determine a delta distance;
generate one or more resolution actions, based at least in part on the first-level features and second-level features, to decrease the delta distance between the one or more supply chain output predictions and the one or more desired supply chain outputs;
adjust the trained probabilistic graphical model based, at least in part, on the one or more supply chain output predictions; and
send instructions to automated machinery, wherein based on the instructions, the automated machinery locates items to add to or remove from an inventory or shipment for the one or more supply chain entities.
9. The system of claim 8 , wherein the computer is further configured to:
generate a Bayesian network graph visualization to display a network graph.
10. The system of claim 8 , wherein the computer is further configured to:
generate recursive level-ranked features based on the list of one or more features and feature classes; and
display on a display device, a network graph comprising the recursive level-ranked features.
11. The system of claim 8 , wherein the computer is further configured to:
receive a query to identify a root cause of a particular metric in the supply chain; and
traverse the probabilistic graphical model to identify features having a largest impact on the particular metric.
12. The system of claim 8 , wherein the one or more supply chain output predictions are based on ensemble data, the ensemble data comprising a combination of the training data and the test data.
13. The system of claim 10 , wherein the ranking for the recursive level-ranked features is based upon a magnitude of change a feature will cause for a KPI or SLA of the supply chain.
14. The system of claim 8 , wherein the historical attributes are normalized, aggregated, or rescaled to allow a direct comparison of data received from different supply chain entities.
15. A non-transitory computer-readable medium embodied with software, the software when executed:
trains a probabilistic graphical model based on historical attributes of a supply chain comprising one or more supply chain entities to represent a performance of one or more supply chain entities in the supply chain;
selects one or more supply chain entity target variables;
collates with a use of one or more machine learning models, a list of one or more features and feature classes pertaining to the one or more selected supply chain entity target variables;
calculates first-level features and second-level features associated with the list of one or more features and feature classes;
generates one or more supply chain output predictions based on the trained probabilistic graphical model, wherein the one or more supply chain output predictions are based on test data;
compares the one or more supply chain output predictions to one or more desired supply chain outputs to determine a delta distance;
generates one or more resolution actions, based at least in part on the first-level features and second-level features, to decrease the delta distance between the one or more supply chain output predictions and the one or more desired supply chain outputs;
adjusts the trained probabilistic graphical model based, at least in part, on the one or more supply chain output predictions; and
sends instructions to automated machinery, wherein based on the instructions, the automated machinery locates items to add to or remove from an inventory or shipment for the one or more supply chain entities.
16. The non-transitory computer-readable medium of claim 15 , wherein the software when executed further:
generates a Bayesian network graph visualization to display a network graph.
17. The non-transitory computer-readable medium of claim 15 , wherein the software when executed further:
generates recursive level-ranked features based on the list of one or more features and feature classes; and
displays on a display device, a network graph comprising the recursive level-ranked features.
18. The non-transitory computer-readable medium of claim 15 , wherein the software when executed further:
receives a query to identify a root cause of a particular metric in the supply chain; and
traverses the probabilistic graphical model to identify features having a largest impact on the particular metric.
19. The non-transitory computer-readable medium of claim 15 , wherein the one or more supply chain output predictions are based on ensemble data, the ensemble data comprising a combination of the training data and the test data.
20. The non-transitory computer-readable medium of claim 15 , wherein the historical attributes are normalized, aggregated, or rescaled to allow a direct comparison of data received from different supply chain entities.