Ksama Arora

Create a basic pipeline in Azure ML Studio - Lab 3

Set Up Compute Resource:

• Go to Azure ML Studio: Navigate to the Azure Machine Learning Studio.
• Create a Compute: Set up a new compute resource.
  • Name the Compute: Provide a name for your compute instance.
  • Choose the Compute Type: Select the appropriate compute type based on your requirements.
  • Configure the Compute: Adjust settings such as the size and region.
  • Create the Compute: Click on the ‘Create’ button to provision the compute resource.

Create a New Classic Pipeline:

• Go to Azure ML Designer: Open the Azure ML Designer (no-code tool).
• Create a New Pipeline: Start a new classic pipeline project.
  • Name Your Pipeline: Provide a name and description for your pipeline.

Data Ingestion:

• Import Datasets: Import the required datasets into your pipeline using Drag and Drop.
  • From Sample Datasets: Select datasets from the component section.
  • Add to Pipeline: Drag and drop the datasets onto the pipeline canvas.
  • Preview Data: Right-click on the datasets and select ‘Preview Data’ to examine the contents.

Data Wrangling:

• Select Columns:
  • Add ‘Select Columns in Dataset’ Module: Drag this module onto the canvas.
  • Configure the Module: Select the columns you need from the dataset.
  • Connect the Module: Link it to the dataset module.

• Join Data:
  • Add ‘Join Data’ Module: Drag this module onto the canvas.
  • Configure the Join: Specify the columns to join on from each dataset.
  • Connect the Datasets: Link both datasets to the ‘Join Data’ module.

• Clean Missing Data:
  • Add ‘Clean Missing Data’ Module: Drag this module onto the canvas.
  • Configure the Module:
  • Cleaning Mode: Select the cleaning mode (e.g., remove entire row or column).
  • Column Options: Specify which columns to clean or apply cleaning to all columns.
  • Connect the Module: Link it to the ‘Join Data’ module.

• Submit Pipeline Job:
  • Select Compute: Choose the compute resource you created earlier.
  • Run the Pipeline: Submit the pipeline job to process the data.

Split Data:

• Add ‘Split Data’ Module: Drag this module onto the canvas.
• Configure the Split:
  • Fraction of Rows in First Output Dataset: Set the fraction to split the data into training and testing sets.
  • Random Seed: Optionally, set a random seed for reproducibility.
• Connect the Module: Link it to the previous module in your pipeline.

Additional Data Wrangling and Modeling (if needed):

• Data Transformation: Add any required transformation modules to preprocess the data.
• Train Model:
  • Add ‘Train Model’ Module: Drag it onto the canvas.
  • Configure the Model: Select the algorithm and set parameters.
  • Connect Training Data: Link the training data from the ‘Split Data’ module.

NOTE IMP:

Components options in Azure ML Designer:

• Preprocess Text: removes stop words from the text
• Split Data Component: splits data into two parts: train and test sets
• Extract N-Gram Features from Text & Convert Word to Vector component: transforms input text into features to make predictions on

• Evaluate Model:
  • Add ‘Evaluate Model’ Module: Drag it onto the canvas.
  • Connect to Trained Model and Test Data: Link the outputs from the ‘Train Model’ and ‘Split Data’ modules.

Run and Monitor the Pipeline:

• Submit the Pipeline: Run the entire pipeline by selecting the compute target.
• Monitor Progress: Track the progress and status of the pipeline run in Azure ML Studio.
• Review Results: Once the run is complete, review the results and metrics.

Publish the Pipeline:

• Set Up Published Pipeline:
  • Select Publish Option: Choose to publish your pipeline.
  • Name and Describe: Provide a name and description for the published pipeline.
  • Pipeline Endpoint: Create or use an existing endpoint for the pipeline.
  • Enable Authentication: Configure authentication and authorization settings.
  • Publish: Finalize and publish the pipeline for use.

By following these steps, you will have created and published a basic machine learning pipeline in Azure ML Studio.