Providing model details

1. Specify model input

Select the type of data that the deployment analyzes and the type of algorithm that you use to build your model. For numeric or categorical data, you must provide information about the training data for your model to configure evaluations. Image models must contain images in the (height) x (width) x (# channels) format, where each point represents either monochrome or RGB values for each pixel. Unstructured text models must include complete text inputs and not a vectorized representation of the text. Fairness and drift evaluations do not support unstructured text or image models.

2. Connect to the training data

Choose the method that you want to use to connect to your training data. If you want Watson OpenScale to connect to training data that is stored in a database or cloud storage, you must select the location and specify connection details. If you want Watson OpenScale to analyze your training data from a notebook to keep the details private, you must run a custom notebook and upload the JSON configuration file that it generates.

3. Select the label column

You must select a single unique feature from the data to serve as the label (prediction) column. This is what the model was designed to predict.

4. Select the training features

Select all the features that were used to train the model before it was deployed. The format of the training data must match the model. For example, if the model expects M and F for the feature Gender, then the training data should have M and F, not Male and Female. Similarly, if a feature column is identified as numeric, this column should be numeric in your model during model training. If a feature column is identified as numeric and you use this column as categorical in your model during model training, then you must update the column to be categorical. For more information, see Managing data for model evaluations.

5. Send model transactions

Send the model request and response data from your deployment to Watson OpenScale to store the transactions in your database. You must send these model transactions to Watson OpenScale with a scoring request that includes measurements of the confidence of predicted outcomes. When you send a scoring request, Watson OpenScale stores the data in a payload logging table within your database and uses the data for model evaluations. For more information, see Managing data for model evaluations.

6. Specify model output

Select a prediction column and a prediction probability column. Watson OpenScale might detect these values for you.

Next steps

Users and roles for model evaluation

Parent topic: Preparing to evaluate a model