Monitor MDM SaaS > Monitoring match model jobs > Monitoring machine learning model jobs
  

Monitoring machine learning model jobs

When you train an ML model, a job runs at each stage of the training process.
When you train a machine learning model, you can view the following job types:

Viewing Match - Start Training jobs

When you start the training process for an ML model, a Match - Start Training job runs to prepare the data for training by generating record pairs.
The following image shows the details for a sample Match - Start Training job:
The image displays the data flow for a match start training process. The page displays the job instance with the overview information.
You can view the following steps in a Match - Start Training job:
  1. 1Generate Tokens. Verifies and ensures that the step generates match keys for all the source records.
  2. 2Clean Up. Cleans up the intermediate states, such as candidate selection criteria and match rules, that the training process generates. The cleanup step is required for a fresh start of the training process after the process has failed.
  3. 3Profile Data. Analyzes the source records and generates internal rules that can provide accurate match results.
  4. 4Generate Candidates. Generates sample match candidates based on the candidate selection criteria of the match model. The process evaluates each candidate record pair against the rules that the data profiling step generates and assigns match scores. The outcome of the step is a list of match candidates with match scores.
  5. 5Generate Seeds. Identifies candidate record pairs that are a match and those that are not a match to generate the seed candidate records pairs.
  6. 6Train and Prepare Batch. Creates a sample ML model trained using the seed candidates and then selects candidate record pairs to prepare the batch for training.

Viewing Match - Calculate Accuracy jobs

After the training data preparation completes, you can start labeling. After you label the record pairs, the Match - Calculate Accuracy job runs to calculate the accuracy of the ML model.
The following image shows the details for a sample Match - Calculate Accuracy job:
The image displays the data flow for a match calculate accuracy process. The page displays the job instance with the overview information.
The Train and Prepare Batch step prepares the record pairs based on the labeling and calculates the accuracy of the ML model.

Viewing Match - Finish Training jobs

After you complete labeling the record pairs, you can finish the training process. When you finish the training process, a Match - Finish Training job runs to complete the training of the ML model.
The following image shows the details for a sample Match - Finish Training job:
The image displays the data flow for a match finish training process. The page displays the job instance with the overview information.
A Match - Finish Training job runs with the following steps:
  1. 1Train and Prepare Batch. Finalizes the record pairs with acceptable accuracy and precision.
  2. 2Save Model. Saves the ML model as part of the match model.

Viewing Match - Delete Model jobs

When you delete a trained match learning model, a Match - Delete Model job runs.
The following image shows the details for a sample Match - Delete Model job:
The image displays the data flow for a match delete model process. The page displays the job instance with the overview information.
You can view the following steps in a Match - Delete Model job: