The visual modeling is to connect the modeling process, configure the parameters and train the models by dragging and dropping the components.

Create Visual Modeling Task

Select [Visual Modeling] from the navigation bar [Model Training], and click "Create" button.

In the pop-up new task, fill in the experiment name (required) and experiment description (optional). Click "Confirm" to create a task.

Copy Visual Modeling Task

Click the [Copy] button in the operation column in the list to copy tasks, as shown in the figure:

Description of Visual Modeling Canvas

Click the task name to enter the visual modeling canvas. There are a lot of different types of components on the left side of the canvas. You can freely explore, drag and drop the components into the canvas. connect them back and forth to complete a complete modeling process.

The right side shows the configurations to be filled for each component. The configurations may include the parameter setting, field setting and resource setting according to different components.

Click the button on the top of the canvas to start or cancel training, and view the historical record.

The historical record displays the historical process of running, including the running time and running status of historical versions. Meanwhile, we can view the historical versions to make comparison of the versions.

The row of buttons below can achieve zoom-in and zoom-out of the canvas, display of actual size, full screen, box selection and opening of thumbnail.

Right click each component to copy and delete components, start execution here, execute here, execute the node, test run a small data volume and view data.

The loading symbol at the right side of the component indicates that the component is running, and the green check mark at the right side indicates that the component running ends.

The buttons at the upper right corner are respectively "Copy Task", Clear Data", "Save" and "Release".

Please note that if you exit the canvas without saving, the operations in the canvas are not recorded.

Click "Data Clearing" to make relevant settings:

After the data are cleared, the middle result data and logs in the running history are cleared.

If one algorithm component and one prediction component run successfully in the canvas, the "Release" button becomes the clickable status. You can release the well-trained models to the model repository by one key for unified management. After clicking "Release", you can fill in the relevant information in the pop-up window.

After the model is released successfully, you can find the model in "Model Repository>User Model".