Skip to main content

Varicent AI Help Center

Classifier

Abstract

Train the model to predict a column with a fixed set of values.

Train the model to predict a column with a set of values.

This tool checks multiple classification algorithms or models and chooses the best one when model_type is set to automatic.

Note

The speed and quality slider is more of a spectrum. This setting determines the number of machine learning models considered. Models that do not support the training data are automatically excluded.

This tool adds two columns to your data: a prediction and a probability. It's the model's certainty about the prediction.

When you run the tool, the data is automatically split: 80% of the data is used for training. The remaining 20% is used for testing. Each model being considered is trained and evaluated to select the one with the best score. This is done 5 times to predict the test values (the 20% of your data). The final score is the average of all 5 scores.

Tip

You can configure this tool without using the configuration menu.

In the Add tool menu, start typing the first few letters of the tool name and press tab to auto-complete. Then start typing the name of the column you want to use and press tab to auto-complete.

Data profile chart visuals

Abstract

From the Row viewer, you can access the Data profile link to open the column details and compare column visuals.

From the Row viewer, you can access the Data profile link to open the column details and compare column visuals. The results are available when you select one of the Classifier, Predictor or Regressor tools.

Accessing Data profile from the Row viewer

  1. In your selected pipe, go to the Row viewer.

  2. Click on the Data profile link in the row viewer.

    The Data profile page opens with column details and compare column visuals.

Note

If you have an explainable tool upstream, you can still get an error message with one of the following issues:

  • The schema has changed in the export. For example, a missing column or an extra column is present.

  • There are multiple explainable tools in the pipe upstream.

  • The pipe changed, and the calculation is now invalid.

If there is no Data profile link, no explainable tool is selected in the upstream pipe.

When to use this tool

Use when you want to predict values.

What is Smart exclude?

Following a successful build using the Classifier tool, Smart exclude identifies and automatically excludes columns that don’t help predict the target column. Smart exclude will only consider columns not already manually excluded. If you want to disable this setting to troubleshoot, test, or run a calculation that is taking too long, go to the Advanced settings under the Configure tab.

How to read the data in this tool

In the row viewer, there are three tabs: Data, Stats and Tool.

The Data tab consists of your imported data. View all of the imported data in one spot.

The Stats tab consists of the statistics for your data. View all of the top values for each column.

The Tool tab visualizes additional insight into the tool and the data. The following columns are available:

  • Accuracy score: Displays the performance measure based on your data set.

  • Column importance: Displays the columns in order of importance.

  • Smart excluded: Displays the columns that don't predict the target column.

Underlying data

In the row viewer, on the Tool tab, there is feature called Underlying data. Click to expand and explore information about your model. The data answers the following questions for your model:

  • How good is your model? This matrix helps explain the strengths and weaknesses of your model.

    how_good_classifier.png

  • How balanced is your model? This section explains how a well-balanced model is more robust and does not have an overly significant impact from a single column. The visual shows how balanced the top 10 columns are.

    how_balanced.png

  • What is my model using to make predictions? This section shows which columns are used to make predictions and the reasons why.

    predictions.png
Configuration

Use the following configuration options to configure the Classifier tool.

Configuring the Classifier tool

  1. Go to the Pipes module from the side navigation bar.

  2. On the Pipes tab, find the pipe you want to work with. Click the pipe to open. For more information about pipes, see the Creating a pipe documentation.

  3. In your pipe, add your data sources.

  4. Click symon_add_icon.png + Tool.

  5. In the Tools modal, search for Classifier in the search bar.

    Tip

    You can also find the Classifier tool in the Learn section.

  6. Click + Add Tool.

  7. Connect the tool to your data set.

  8. In the configuration pane, enter the following information:

    Table 65. Classifier tool configuration

    Field

    Description

    Train type

    Select the train type that you want to use to train your data.

    Target column

    Select the target column to train your data against.

    Advanced

    Performance Measure

    Select the type of performance measure to use:

    • Automatic: Let Varicent AI make the decision for you.

    • Accuracy: How often the model is correct about which class it predicts.

    • Area under the curve: The area under the curve (AUC) is the measure of the model's ability to separate two values. The higher the value, the greater the concentration of positive case in the top scores that the model produces.

    • Precision: How often an observation which is predicted to be this value has this value in reality.

    • Recall: How often the model correctly predicts observations with a given value.

    Speed versus Quality slider

    Use the slider to indicate if you want speed versus quality when the Classifier is working.

    Exclude columns

    Select the column(s) that you want to exclude from the Classifier.

    Smart exclude

    elect this option if you want to have Smart Exclude identify and automatically exclude columns that don’t help predict the target column after you build.