Workflow

Updated at:
Copy as MD

This topic describes the business collaboration flow and product usage flow for Multi-Party Secure Modeling.

Business collaboration flow

The business collaboration flow includes an offline communication procedure and a platform participation procedure. For details about the business collaboration flow, see the following flowchart.

1636433494754_EE65DF4D-1B18-46f9-8344-65F41CDF8E8C.png

Product usage flow

The Multi-Party Secure Modeling service enables you to complete model training by exchanging information while ensuring that the raw data of all parties remains on-premises. After model training is complete, each participant holds only a part of the trained model. During the prediction phase, each participant uses their part of the model and their data to perform a joint prediction and obtain the final scoring result.

The main product usage flow is as follows:

2

The Multi-Party Secure Modeling console provides end-to-end features for joint modeling business applications, including model development, model deployment, and invocation tracking. The features are as follows:

  • Model development

    The model development phase uses the privacy-preserving set intersection and model training modules in the Multi-Party Secure Modeling console. After model training is complete, you can generate and view a model report in the console. After these operations are complete, you can proceed to the model review stage. For more information about the model development flow, see Model development overview.

  • Model deployment

    After the model passes the review, you can deploy it. The model deployment phase uses the model publishing and service integration modules in the Multi-Party Secure Modeling console. After model deployment is complete, you can proceed to the service invocation stage. For more information about the model deployment flow, see Model deployment overview.

  • Invocation tracking

    During the service invocation process, the Multi-Party Secure Modeling console provides a feature to monitor model invocations. For more information about the invocation tracking flow, see Service monitoring overview.