Create a machine learning model

Machine Learning Model Creation enables the creation of a machine learning (ML) model using a CSV dataset for value predictions.

Create machine learning (ML) functionality

Use this procedure to create a machine learning (ML) model in PreDAI.

1. On the Admin tab, in the AI group, select PreDAI to display the Predictive Engineering AI window.

2. Select an ML Model World from the dropdown list.

3. Enter the ML model name.

4. Enter the ML model description.

5. Select Refresh

   The ML Model World updates.

6. Select View Existing Model.

   You can view or delete existing models.

   

   The ML model is created and available in the ML Model Worlds.

Input data for ML model

7. Select Import CSV to display the DataTypeForm window.

   

8. In the Set Type and Purpose grid, select Feature or Label.

9. Select a Datatype, which is optional.

10. Select Apply.

    

    Note:
    The CSV header must be alphanumeric and must start with an alphabetical character or an underscore (_)
    The Purpose option displays Feature by default.

Select ML task

In Select ML task choose Classification or Regression.

Select training algorithm

Automatic Algorithm Selection

Automatic Algorithm Selection enables the application to evaluate multiple algorithms and train a model to automatically achieve the best performance.

The grid provides multiple modes, each with its own checkbox and approach to training optimization.

AutoML: Simplifies training by automating model selection and optimization.

Tune Hyperparameters: Improves model performance through automatic hyperparameter tuning.

Manual: Maintains full control over model configuration and training.

AutoML

When you select the AutoML training option, the Time to Train field is enabled.

Purpose of Time to Train

The Time to Train setting specifies the duration AutoML has to complete the training process. It does the following:

• Evaluates multiple machine learning algorithms.

• Tests different hyperparameter combinations.

• Executes the maximum number of training trials permitted by the specified time limit.

Training Process

• AutoML executes trials continuously until the allocated time expires.

• Each trial represents a unique combination of algorithm and hyperparameters.

• AutoML completes an in-progress trial before stopping when the time limit is reached.

• AutoML selects the model with the best performance metric once all possible trials are complete.

• It automatically saves the best-performing model to the database.

Insufficient Time Scenario

If the specified time is too short to complete a training trial, AutoML stops and displays an error message in the User Messages window. Ensure sufficient training time is allocated before starting the process to avoid this error.

Tune HyperParameters

The Tune Hyperparameters option enables you to select a specific machine learning algorithm for training.

Note: Since AutoML automatically performs hyperparameter optimization in this mode, the Edit Hyperparameters Settings button is disabled.

Functionality

• Select the desired algorithm.

• Multiple hyperparameter combinations are automatically evaluated for the selected algorithm.

• Training trials are executed across different hyperparameter configurations.

• The best-performing model is identified and automatically saved to the database.

• If a trial is in progress when the specified time is reached, AutoML completes the trial before stopping.

Time Allocation

Hyperparameter tuning runs for the duration specified in the Time to Train setting.

The system performs the maximum number of tuning trials permitted by the allocated time.

Once the time limit is reached, the best-performing model identified during the tuning process is finalized and saved.

Note: It is recommended that you allocate sufficient training time to ensure effective hyperparameter optimization and improved model performance.

Insufficient Time Scenario

If the specified time is too short to complete a training trial, AutoML stops and displays an error message in the User Messages window.

Manual

Manual provides full control over model configuration and training.

Functionality

• Selects the desired machine learning algorithm.

• Hyperparameters can be configured manually using the Hyperparameter Settings window.

• The model is trained on defined specifications.

• No automated model selection or hyperparameter tuning is performed in this mode.

• Upon completion, the trained model is saved to the database.

  Note: This option is recommended if you require full control over algorithm selection and hyperparameter configuration.

Input Data Requirements

The input CSV file must not contain any empty or null values.

Incomplete or missing data may produce undetermined or inaccurate results.

Ensure the dataset is clean and complete before initiating model training.

This requirement is consistent with previous versions of No-Code ML.

Proper data preparation is essential for reliable and accurate machine learning results.