Advanced application of model operations in energy
A method implements advanced application of model operations in energy. The method includes presenting an inspection view with a model list. The method further includes receiving a selection of a list entry, corresponding to a model, from the model list, wherein the model of the list entry is a champion model selected from multiple challenger models corresponding to the list entry. The method further receiving deployment settings of the model, presenting a deployment view with a deployment list with the model, and updating a deployment status of the model in the deployment view.
1 . A method, comprising:
presenting an inspection view with a model list, the model list comprising a plurality of list entries, each list entry corresponding to a respective data analysis project, each list entry comprising:
a plurality of challenger models corresponding to the respective data analysis project for the list entry; and
a selected model comprising a champion model for the respective data analysis project, the champion model being automatically selected from among the plurality of challenger models periodically after retraining of the plurality of challenger models is performed;
receiving a selection of a selected list entry from among the plurality of list entries in the model list;
receiving deployment settings of the selected model of the selected list entry;
presenting a deployment view with a deployment list with the selected model; and
updating a deployment status of the selected model in the deployment view,
wherein the retraining of the plurality of challenger models is performed periodically after a deployment of the selected model based on at least a performance metric monitored during the deployment.
2 . The method of claim 1 , further comprising:
storing a model lineage, of the selected model, identifying a version of the selected model, a training time of the selected model, training data the selected model was trained with, metadata of the selected model, algorithms of the selected model, parameters of the selected model, and validation accuracies of the selected model.
3 . The method of claim 1 , further comprising:
for each list entry in the model list, designating the champion model over the plurality of challenger models based on a metric corresponding to the respective data analysis project for the list entry.
4 . The method of claim 1 , further comprising:
presenting the list entry with a validation accuracy of the selected model.
5 . The method of claim 1 , further comprising:
presenting the plurality of challenger models with corresponding validation accuracies.
6 . The method of claim 1 , further comprising:
deploying the selected model using the deployment settings and a procedure call standard selected from among a plurality of procedure call standards.
7 . The method of claim 1 , further comprising:
deploying the selected model using continuous deployment and continuous delivery using an automated pipeline to train, test, deploy, and continuously improve the selected model in production.
8 . The method of claim 1 , further comprising:
monitoring deployment of the selected model with performance metrics, audit logs, and relevant performance indicators.
9 . The method of claim 1 , further comprising:
identifying degradation of the selected model over time using a model drift alert.
10 . The method of claim 1 , further comprising:
monitoring deployment of the selected model to auto-scale compute resources allocated to the selected model.
11 . The method of claim 1 , further comprising:
presenting the deployment status of the selected model with a link to the selected model.
12 . The method of claim 1 , wherein the deployment status comprises one of “pass”, “fail”, and “in progress”.
13 . The method of claim 1 , wherein the retraining of the plurality of challenger models is performed periodically on a schedule of at least one day between each retraining.
14 . The method of claim 1 , wherein the champion model is automatically selected from among the plurality of challenger models based on one or more model accuracy metrics that identify an accuracy of each model, the one or more model accuracy metrics comprising one or more of: a mean absolute value (MAV), a mean absolute error (MAE), a root mean squared error (RMS), or a Pearson correlation coefficient (PCC).
15 . The method of claim 1 , wherein the deployment status is indicated by displaying, on a user interface, at least two of: an image, text, or a link.
16 . The method of claim 15 , wherein, responsive to the deployment status being “pass”, the link is displayed on the user interface.
17 . A system, comprising:
an inspection view;
a deployment view; and
a server application executing on one or more servers and configured for:
presenting the inspection view with a model list, the model list comprising a plurality of list entries, each list entry corresponding to a respective data analysis project, each list entry comprising:
a plurality of challenger models corresponding to the respective data analysis project for the list entry; and
a selected model comprising a champion model for the respective data analysis project, the champion model being automatically selected from among the plurality of challenger models periodically after retraining of the plurality of challenger models is performed;
receiving a selection of a selected list entry from among the plurality of list entries in the model list;
receiving deployment settings of the selected model of the selected list entry;
presenting the deployment view with a deployment list with the selected model; and
updating a deployment status of the selected model in the deployment view,
wherein the retraining of the plurality of challenger models is performed periodically after a deployment of the selected model based on at least a performance metric monitored during the deployment.
18 . The system of claim 17 , wherein the server application is further configured for:
storing a model lineage, of the selected model, identifying a version of the selected model, a training time of the selected model, training data the selected model was trained with, metadata of the selected model, algorithms of the selected model, parameters of the selected model, and validation accuracies of the selected model.
19 . The system of claim 17 , wherein the server application is further configured for:
for each list entry in the model list, designating the champion model over the plurality of challenger models based on a metric corresponding to the respective data analysis project for the list entry.
20 . The system of claim 17 , wherein the server application is further configured for:
presenting the list entry with a validation accuracy of the selected model.
21 . The system of claim 17 , wherein the server application is further configured for:
presenting the plurality of challenger models with corresponding validation accuracies.
22 . The system of claim 17 , wherein the server application is further configured for:
deploying the selected model using the deployment settings and a procedure call standard selected from among a plurality of procedure call standards.
23 . The system of claim 17 , wherein the server application is further configured for:
deploying the selected model using continuous deployment and continuous delivery using an automated pipeline to train, test, deploy, and continuously improve the selected model in production.
24 . A method, comprising:
displaying, on a user interface, an inspection view with a list of models, the list of models comprising a plurality of list entries, each list entry corresponding to a respective data analysis project, each list entry comprising:
a plurality of challenger models corresponding to the respective data analysis project for the list entry; and
a selected model comprising a champion model for the respective data analysis project, the champion model being automatically selected from among the plurality of challenger models periodically after retraining of the plurality of challenger models is performed;
receiving a selection of a selected list entry from among the plurality of list entries in the list of models;
receiving deployment settings of the selected model of the selected list entry;
displaying, on a user interface, a deployment view with the list of models; and
updating a deployment status of the selected model in the deployment view,
wherein the retraining of the plurality of challenger models is performed periodically after a deployment of the selected model based on at least a performance metric monitored during the deployment.