> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.umnai.com/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.umnai.com/_mcp/server.

# Deployed model views

Deployed model views generate structured outputs from deployed models.

Use them after deployment to request predictions, inspect explanations, diagnose model behaviour, or evaluate performance.

## Views by concern

### Prediction results

[**Results**](./results-view) returns the prediction output together with the query input and the underlying dataframes used to explain the prediction.

Use it when you need the complete inference result, or when you want the raw prediction and explanation data in one response.

### Attribution and rules

[**Feature attribution**](./feature-attribution-view) explains a prediction in terms of input features.

[**Module attribution**](./module-attribution-view) explains a prediction in terms of model modules, including the partitions and rules that contributed to the result.

[**Decision actions**](./decision-action-table-view) returns a partition-level table for inspecting activations, attributions, conditions, and rule expressions.

### Model behaviour

[**Module dependency**](./module-dependency-view) shows how selected module features relate to attribution values.

Use it when you need to understand how a module behaves across feature values, rather than how a single prediction was attributed.

### Diagnostics

[**Anomalies**](./anomaly-view) identifies anomalous bins in the data used by the model.

[**Strengths and weaknesses**](./saw-view) estimates prediction strength, uncertainty, and adjusted error when target observations are available.

### Evaluation

[**Classification evaluation**](./classification-evaluation-view) evaluates classification performance with threshold and global metrics.

[**Regression evaluation**](./regression-evaluation-view) evaluates regression performance with error and fit metrics.