Classifier
Train the model to predict a column with a fixed set of values.
Train the model to predict a column with a fixed set of values.
This tool looks at all the available classifier tools and picks the best one.
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. The probability is the likelihood that the prediction is accurate.
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 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 visuals
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.
In your selected pipe go to the Build tab and select the classifier node in a built pipe.
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 profle link, there is no explainable tool selected in the pipe upstream.
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 is a visualization of additional insight into the tool and the data. The following columns are available:
Accuracy score: Displays the score based on your data set.
Advanced score: Displays a baseline average and the actual score for accuracy, precision, recall, F-1, and area under ROC curve.
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 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.
What is my model using to make predictions? This section shows which columns are used to make predictions and the reasons why.
