Step Four: Predict and Execute Capturing Results in AIML Prediction Run UI

Prediction Run UI

In Prediction Run UI, each prediction run is defined by the training run that has been used and other parameters that will vary by the metric. For example, Order Cycle Time, it is the order filter criteria.

Each time a metric invokes the ‘Predict’ button, the results will be stored here. Additionally, each scheduled job run for the job type 'Generate Prediction Run' will result in an entry here.

The following fields are in the Prediction Run UI:

Field Name:	Description:
Prediction Run Nbr	System generated number on invoking Predict action or on running a 'Generate Prediction Run' job.
Training Run Nbr	Reference to the training Run from the AIML Training Models
Prediction Parameters	Reference to the parameters used to execute the model against new data to generate predictions
Default flag	Copied from training run; there can be multiple runs for same metric with this enabled. the latest one is the relevant one.
Status	created, training, completed, failed, cancelled

There is a drill down detail button available when a user selects any given record. This shows the output data for that run. It is specific to the metric of the run that the user selects. The output will be one of the following KPIs:

- - Intelligent Cycle Count
  - Order Cycle time
  - Order Waiting Time
  - Order processing Time

Users are also able to use search/filters for Prediction Run Number, Training Run Number, Status and Default Flag. Each drill down screen also supports CSV download and appropriate filters as needed.

Order Cycle Time, Order Waiting Time, Order Processing Time, and Intelligent Cycle Count

When selecting a record in the AI/ML Models screen for any of the metrics, a user must click the ‘Predict’ button to carry out the prediction. A popup will be displayed to further finetune the models.

Note: If no field is populated, then a user will be shown an error.

On clicking 'OK', the execution i.e. the prediction for the respective metrics will commence for the set of orders selected. (On 'Cancel', the action is cancelled.)
On completing the execution, a record is generated in the Prediction Run (Inquiry) UI which records the parameters and details used to run that prediction.
Also, the execution will generate the output i.e. the Order Cycle Time or Order Waiting Time or Order Processing Time depending on the metric of the record selected for prediction, viewed in a separate UI screen ex: 'AIML Order Cycle Wait Processing Time

AIML Models Inquiry

Every training template execution, whether manual or scheduled will result in an entry here. For metrics that produce a model, the model generated will be stored within this screen.

The following fields are supported within this UI:

Field Name	Description
Facility	User's context facility
Training Run Number	System generated number on running a AIML training template
AIML Training Template	name of the training template ran for the specific run number
Training Start Date	Start date of the training data
Training End Date	End date of the training data
AIML Algorithm	Algorithm for which the model was generated
Training Filters	Copied from template
Algorithm Parameters	Copied from template
Default Flag	Same as from template
Status	Created, training, completed, tailed, cancelled
Message Text	Shows the latest message from the training run Will be a reference number or link updated when training completes. some metrics won’t have a model (like basket analysis) This should be copied back to the template "Last Trained Model"
AIML Model Reference	Shows the performance of the model
Performance Metrics	Mean absolute error, r^2
Create Timestamp	Date and time of the model/record. Creation
Create User	User who generated the model/record (one who ran the AI/ML training template)
Metric	Metric being training, comes from template

Action Buttons in AIML Models Inquiry:

- - Logs
  - Predict
  - Prediction Run

Each of the drill down screens supports CSV download and appropriate filters as needed.