Health component recommendation setup

Data replication to Data Lake is required to provide the dataset for the health component recommendation feature.

The model analyzes incident data and associated health component classifications to generate recommendations.

Replicating structured and unstructured incident data ensures consistent, historical, and organization-specific inputs for training and inference.

The feature uses Landmark Replication Sets to publish required business class data. The delivered replication set HRT_AI includes these required business classes:

• Incident
• IncidentResourceHealthComponent

These business classes require these fields:

• Incident: HROrganization, Description, Incident, IncidentSummary
• IncidentResourceHealthComponent: HROrganization, Incident, OccupationalHealthComponent

The replication set must include an imported payload file.

See Infor Landmark Technology Library (Cloud) and select Administration > System Administration > Data replication > Creating and running replication sets > Starting and scheduling data replication.

After replication, you must validate the data in Infor Data Lake by using these table names:

• HRT_AI_Incident
• HRT_AI_IncidentResourceHealthComponent

See Infor OS User and Administration Library (Cloud) and select User > Data Fabric > Viewing data in Data Lake.

Datasets in the AI Platform provide structured input for training and prediction. Datasets are created from replicated Data Lake tables and a payload dataset for real-time inference.

These datasets provide the model with historical incident data and input text for classification.

These Data Lake tables are required:

• HRT_AI_Incident
• HRT_AI_IncidentResourceHealthComponent

See Infor Artificial Intelligence User Guide (Cloud) and select Content deployment and configuration > Adding the dataset.

The custom algorithm defines the logic that trains the health component recommendation model. The custom algorithm processes incident text data and maps it to predefined Occupational Health Components.

These components are required:

• Custom algorithm code package, for example, HCM_RecommendHealthComponents_CA
• Hyperparameter configuration, if applicable

The algorithm must support training by using input text fields from datasets and associated health component labels for classification.

See Infor Artificial Intelligence User Guide (Cloud) and select Machine learning > Machine learning custom algorithms > Uploading a source file.

The quest defines the machine learning pipeline that trains and evaluates the model. The quest connects datasets, applies transformations, runs the algorithm, and produces a trained model.

The production quest supports real-time predictions through an exposed endpoint.

These components are required:

• Training datasets: Incident and IncidentResourceHealthComponent
• Payload dataset
• Trained model artifact
• Production quest configuration that includes an ingest dataset (payload) and predictor node
• Endpoint name for integration with HR Talent

See Infor Artificial Intelligence User Guide (Cloud) and select Content deployment and configuration > Creating the quest and deploying the endpoint.

The workflow automates model retraining and endpoint updates by using refreshed data. The workflow ensures that the recommendation model reflects the latest incident records in Data Lake.

These configuration elements are required:

• Workflow template
• Schedule configuration for execution frequency
• Dataset refresh trigger
• Model retraining execution
• Production quest update
• Endpoint redeployment

Scheduling must align with the data replication frequency and data availability.

See Infor Artificial Intelligence User Guide (Cloud) and select Optimization > Model maintenance > Creating a workflow.

See Infor Artificial Intelligence User Guide (Cloud) and select Optimization > Model maintenance > Setting up a workflow schedule.