![](https://crypto4nerd.com/wp-content/uploads/2023/11/0S5TQmtgS3VDtGpkU-1024x590.png)
Let’s start with the Azure ML Designer Model Building.
Create Azure ML Job
Once the data asset is created, we can create the model inside the Azure ML Designer tool to train the model.
- Click on the ‘Designer’ tab.
- Click on ‘+’ Create a new pipeline using classic prebuilt components.
- Under the ‘Asset Library’ tab select the ‘Data’ tab(From Data and Component), then select the data asset name (e.g. customer-transaction in my case).
4. Create the basic pipeline to check how many features are important from available features.
- Click on the ‘Component’ Tab.
a. Search, Drag & Drop: “Select Columns in Dataset”
- Connect the component to the dataset.
- Double-click the component.
- Click on ‘Edit column’.
- Follow steps given in image to DROP UNWANTED COLUMNS. Note: To drop columns use Exclude and to select useful columns use Include.
b. Search, Drag & Drop: “Filter Based Feature Selection”
- Connect the component to the ‘Select Columns in Dataset’ component.
- Double-click the component.
- Provide “Number of desired features” (will select the top n features that are affecting the target column).
- Provide a ‘Target’ column by clicking on ‘Edit Column’. Note: Perform this step after analyzing data carefully as it can drop certain important features that may affect the accuracy of the model.
c. Search, Drag & Drop: “Split data”
- Connect the node ‘Filtered dataset’ of the “Filter Based Feature Selection” container to the ‘Dataset’ of the ‘Split Data’ Container.
- Double-click the component.
d. Search, Drag & Drop: “Two-Class Logistic Regression” Algorithm for train model
e. Search, Drag & Drop: “Train Model”
- Connect the node ‘Untrained model’ of the “Two-Class Logistic Regression” container to the ‘Untrained model’ of the ‘Train Model’ Container.
- Connect the node ‘Result Dataset1’ of the “Split data” container to the ‘Dataset’ of the ‘Train Model’ Container.
- Double click ‘Train Model’.
- Click on “Edit column”.
- Select/Type the name of the target column.
f. Search, Drag & Drop: “Score Model”
- Connect the node ‘Trained model’ of the “Train Model” container to the ‘Trained model’ of the ‘Score Model’ Container.
- Connect the node ‘Result Dataset2’ of the “Split data” container to the ‘Dataset’ of ‘Score Model’ Container.
g. Search, Drag & Drop: “Evaluate Model”
- Connect the node ‘Scored Dataset’ of the “Score Model” container to the ‘Scored Dataset’ of the ‘Evaluate Model’ Container.
5. Click on ‘Settings’ > “Select Compute cluster”. > Save
6. Click on the Validate
button. (Check for errors)
7. Click on the Submit
button
Note: Check ‘Select existing experiment’ if exists. This window will create a New Experiment and will create a Pipeline Job under that experiment. To check the Job details click on ‘Jobs’ under the ‘Assets’ tab Select the name of your experiment under the ‘Experiment’ tab and click on the Job name.
Jobs > All experiments -> name_of_your_Experiment -> name_of_your_Job