AutoML Builder

Contents

◷ Reading Time: 7 minutes

The AutoML capability of FlexRule automatically examines different machine learning algorithms, finds the best algorithm, and generates the best model to make a prediction.

It can create the following three types of models.

Regression
Multiclass Classification
Binary Classification

Building the Model

You can find the AutoML Builder under Decision Analytics.

The method for all the builders is the same. Therefore, we will continue the Regression model builder.

Once you install the Analytics extension (FlexRule.Extensions.Analytics), go to,

Add New Document --> Decision Analytics --> Regression

Then it will launch the Regression ML Model Builder.

There are three steps:

Dataset: To upload the dataset
Input and Outputs: Select the input and prediction columns
Training Process: Train the model

Dataset

In the first step, we will be uploading the dataset to start building the model.

Select the dataset in CSV file type.

The dataset should have a minimum of 100 rows.

Once the dataset is uploaded successfully, you will see the file information.

Size: Size of the uploaded file
Separator: Delimiter of the CSV file
Number of Rows: The number of rows in the uploaded dataset
Number of Columns: Number of columns in the uploaded dataset

You will also have the option to preview the uploaded dataset.

Click Next Step to go to the next step.

Inputs and Outputs

In this step, we will be selecting the inputs and the output column to make predictions.

The input/ output lists are auto-generated from the file uploaded in the previous step.

Select the output column from the drop-down list.

Select the input columns. The prediction of the above-selected output column is determined based on the input columns you select in the step. Therefore, if there are any columns that you don’t need to be used in the prediction, you can un-tick them here.

Click Next Step to go to the next step.

Training Process

This is the final step that allows you to choose the options for training.

Time to Train (Seconds): You can choose how many seconds you would like to spend to train the model.
Optimizing Metric: Determines the metric to be optimized during model training.
For information on each metric:
Algorithm: List of algorithms to be chosen from.
Permutation Count: Number of permutations.
Path to Store ML Model: By default, it will save in the project folder. But you can choose a preferred path.
Logic Documents: This option cannot be changed. It will create a DRD document at the end of the training.
Artifacts Home: By default, it is the project’s root folder.

Finally, you can start training the model by clicking on Start.

Once the process starts, you can see the steps in the Background Tasks window.

When the training is completed successfully, you will be able to see the completed steps.

Click on the completed task and click View Task Result.

Then you will see the report of the built model.

Automatically Created Model

You will also see the automatically created files.

These are the files:

<CSV file name>.DRD.xml:
The decision model can be used to test the model and integrate into your project.
<CSV file name>.InputOutput.Concept.xml:
The concept file defines the inputs and output (prediction).
<CSV file name>.frm:
Model Building Report containing details such as the chosen best algorithm, how many algorithms explored, etc.

<CSV file name>.DRD.xml