This topic describes how to create a pipeline in Machine Learning Platform for AI (PAI).

Machine Learning Designer allows you to create a pipeline by using a template or create a custom pipeline.

Machine Learning Designer provides demos to show you how to create a custom pipeline and create a pipeline by using a template.
  • Demo for a custom pipeline

    Machine Learning Designer provides more than one hundred algorithm components that you can use to create custom pipelines and supports access to a variety of data sources, such as MaxCompute and Object Storage Service (OSS). This improves your modeling efficiency. The following procedure provides an example on how to create a custom pipeline:

    1. Create a custom pipeline

      Upload raw data to MaxCompute or OSS, and configure the data source for the pipeline.

    2. Prepare and preprocess data

      Preprocess the raw data to generate a model training set and a model prediction set.

    3. Data visualization

      Process the source data or the intermediate results in a visualized manner to obtain data analysis results.

    4. Generate a model

      Create a model by using the algorithm components that meet your business requirements and the model training set.

    5. Evaluate a model

      Make predictions by using the trained model and the model prediction set. Evaluate the quality of the model based on the correct answers in the model prediction set.

  • Demo for a pipeline created by using a template

    You can create a pipeline by using a template. After the pipeline is created and works as expected, you can deploy the model. For more information about the demo, see Create experiments based on templates.