Sample 5: Train, Test, Evaluate for Binary Classification: Adult Dataset

By for November 26, 2014
This experiment demonstrates how we can build a binary classification model to predict income levels of adult individuals. The process includes training, testing and evaluating the model on the Adult dataset.
#Binary Classification: Income Level Prediction In this sample experiment we will train a binary classifier on the [Adult]( dataset, to predict whether an individual’s income is greater or less than $50,000. We will show how you can perform basic data processing operations, split the dataset into training and test sets, train the model, score the test dataset, and evaluate the predictions. ##Creating the Experiment## 1. Drag and drop the **`Adult Census Income Binary Classification dataset`** module into your experiment's workspace. 2. Add a **Clean Missing Data** module, and use the default settings, to replace missing values with zeros. Connect the dataset module output to the input port. 3. Add a **Project Columns** module, and connect the output of **Clean Missing Data** module to the input port. 4. Use the column selector to exclude these columns: `workclass`, `occupation`, and `native-country`. We are excluding these columns because we don't want their values to be used in the training process. By default, Azure ML Studio treats all columns as features except for the target variable (the Label column). Alternatively, you could use the **Metadata Editor** module, select the excluded columns, and then choose _ClearFeatures_ from the **Fields** dropdown list. ![Select Columns]( 5. Add a **Split** module to create the testing and test sets. Set the *Fraction of rows in the first output dataset* to 0.7. This means that 70% of the data will be output to the left port and the rest to the right port of this module. We will use the left dataset for training and the right one for testing. 5. Add a **Two-Class Boosted Decision Tree** module to initialize a boosted decision tree classifier. 6. Add a **Train Model** module and connect the classifier (step 5) and the training set (left output port of the **Split** module) to the left and right input ports respectively. This module will perform the training of the classifier. 7. Add a **Score Model** module and connect the trained model and the test set (right port of the **Split** module). This module will make the predictions. You can click on its output port to see the actual predictions and the positive class probabilities. 8. Add an **Evaluate Model** module and connect the scored dataset to the left input port. To see the evaluation results, click on the output port of the **Evaluate Model** module and select *Visualize*. ![]( ##Results From these results, you can see that the **Two-Class Boosted Decision Tree** is fairly accurate in predicting income for the `Adult Census Income` dataset. ![results](