---
schema_version: "1"
id: d7cc3f921f2fe686dfd592cd150ee2a547a947decbb1b6b272f016b33d1304e5
title: Overall Performance Tab for Multiclass Classification Use Cases
collection: admin
version: current
locale: en-us
portal_url: https://docs.salesforce.com/admin/current/en-us/analytics/264-0-0/bi_edd_model_metrics_overall_multiclass.md
canonical_url: https://help.salesforce.com/s/articleView?id=analytics.bi_edd_model_metrics_overall_multiclass.htm&release=264.0.0&type=5
source_site: help.salesforce.com
content_type: documentation
content_hash: 7b6f814c55d91a6cbc899b2511fb56e1a671d5fcfd648f43cca13c0dc16f3db4
description: The Overall Performance tab shows key metrics for model quality.
products:
  - CRM Analytics
  - Cross Cloud
release: "264"
---
# Overall Performance Tab for Multiclass Classification Use Cases

The Overall Performance tab shows key metrics for model quality.

![Note](https://sf-zdocs-cdn-prod.zoominsoftware.com/tdta-analytics-bi-264-0-0-production-enus/ec791473-f5b4-4f64-a656-9104f455d966/images/icon_note.png)

Note Einstein Discovery stories are now models. We wish we could snap our fingers to update the name everywhere, but you can expect to see the previous name in a few places until we replace it.

## Navigate to the Overall Performance Tab

In Performance, click **Model Evaluation**, then click **Overall Performance**.

![Overall Performance tab showing accuracy analysis results](https://sf-zdocs-cdn-prod.zoominsoftware.com/tdta-analytics-bi-264-0-0-production-enus/ec791473-f5b4-4f64-a656-9104f455d966/bi/images/bi/bi_edd_model_metrics_overall_multiclass.png)

## Class Selector

Select a class to see its performance statistics.

## ROC Curve

Receiver Operating Characteristic Curve. Displays the performance measurement for the selected class. ROC is a probability curve and AUC (Area Under the Curve) represents the degree or measure of separability. This chart shows how well the model is able to distinguish between classes.

- Y-axis: True Positive Rate: TPR = TP / (TP + FN)
- X-axis: False Positive Rate>: FPR = FP / (FP + TN)
- Model (blue line)
- No Model (gray line)—the same as random chance

## Row Count for the Selected Class

Number of observations in the training dataset associated with the selected class.

## AUC

Area Under the Curve. Measures the model’s rate of correct classification for the selected class.

Range:

- 0.5 means that the model performed no better than random guessing.
- 1.0 means that the model correctly classified the data 100% of the time. An AUC of 1.0 is suspect because it can indicate data leakage: the data used to train your model includes one or more columns that contain the information that you are trying to predict.

## Confusion Matrix for the Selected Class

The confusion matrix is used to evaluate the trade-offs between different error types for the selected class. It displays how many times the model correctly and incorrectly classifies an observation as true and false for the class. Key Metrics:

<table class="slds-table slds-table_bordered slds-max-medium-table_stacked slds-m-bottom_small ">
        <colgroup>
          <col style="width:14.285714285714285%">
          <col style="width:85.71428571428571%">
        </colgroup>
        <thead class="slds-text-title_caps ">
          <tr class="">
            <th scope="col" class="slds-cell-wrap" style="vertical-align:top;" id="d163365e96">Metric</th>
            <th scope="col" class="slds-cell-wrap" style="vertical-align:top;" id="d163365e98">Description</th>
          </tr>
        </thead>
        <tbody>
          <tr class="">
            <td data-label="Metric" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e96 "><strong class="uicontrol">F1 Score</strong></td>
            <td data-label="Description" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e98 ">
              <p>The <em>F1 score</em> represents the harmonic average
                of precision and recall.</p>
              <p>Formula:</p>
              <p>F1 = 2
                * ( (precision * recall) / (precision + recall)
                )</p>
              <p>where</p>
              <ul>
                <li>precision represents the number of
                  correct positive results divided by the number of
                  all positive results.</li>
                <li>recall represents the number of correct
                  positive results divided by the number of all
                  relevant observations (all observations that have
                  been identified as positive).</li>
              </ul>
            </td>
          </tr>
          <tr class="">
            <td data-label="Metric" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e96 "><strong class="uicontrol">True Positive Rate</strong></td>
            <td data-label="Description" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e98 ">
              <p>The <em>True Positive Rate</em> (<em>TPR</em>) measures
                the proportion of true positives that the model
                predicted correctly.</p>
              <p>Formula:</p>
              <p>TPR = TP /
                (TP + FN)</p>
              <p>where</p>
              <ul>
                <li>TP represents the number of true
                  positives (positive prediction with a positive
                  result)</li>
                <li>FN represents the number of false
                  negatives (negative prediction with a positive
                  result)</li>
              </ul>
              <p>Range:</p>
              <ul>
                <li>one (1) represents a perfect test (100% true
                  positives)</li>
                <li>zero (0) represents the worst possible value
                  (0% true positives).</li>
              </ul>
              <p style="margin-bottom:0;">The True Positive Rate is also known as
                <em>Sensitivity</em> or <em>Recall</em>.
              </p>
            </td>
          </tr>
          <tr class="">
            <td data-label="Metric" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e96 "><strong class="uicontrol">Precision</strong></td>
            <td data-label="Description" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e98 ">
              <p><em>Precision</em> describes the proportion of
                positive results that are true positive results.
              </p>
              <p>Formula:</p>
              <p>Precision = TP / (TP +
                FP)</p>
              <p>where</p>
              <ul>
                <li>TP represents the number of true positives
                  (positive prediction with a positive result)</li>
                <li>FP represents the number of false positives
                  (positive prediction with a negative result)</li>
              </ul>
              <p>Range: Precision is also known as
                <em>Positive Predictive Value</em> (<em>PPV</em>).
              </p>
              <ul>
                <li>1 represents a perfect test (100% true
                  positives)</li>
                <li>0 represents the worst possible value (0% true
                  positives)</li>
              </ul>
            </td>
          </tr>
          <tr class="">
            <td data-label="Metric" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e96 "><strong class="uicontrol">Negative Predictive
                Value</strong></td>
            <td data-label="Description" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e98 ">
              <p><em>Negative Predictive Value</em> (NPV) describes
                the proportion of negative results that are true
                negative results.</p>
              <p>Formula:</p>
              <p>NPV = TN / (TN
                + FN)</p>
              <p>where</p>
              <ul>
                <li>TN represents the number of true
                  negatives (negative prediction with a negative
                  result)</li>
                <li>FN represents the number of false
                  negatives (negative prediction with a positive
                  result)</li>
              </ul>
              <p>Range:</p>
              <ul>
                <li>1 represents a perfect test (100%)</li>
                <li>0 represents the worst possible value</li>
              </ul>
            </td>
          </tr>
          <tr class="">
            <td data-label="Metric" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e96 "><strong class="uicontrol">Markedness</strong></td>
            <td data-label="Description" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e98 ">
              <p><em>Markedness</em> measures the trustworthiness of
                positive and negative predictions by the
                model.</p>
              <p>Formula:</p>
              <p>Markedness = PPV + NPV
                - 1</p>
              <p>where</p>
              <ul>
                <li>PPV is the Positive Predictive
                  Value</li>
                <li>NPV is the Negative Predictive
                  Value</li>
              </ul>
            </td>
          </tr>
          <tr class="">
            <td data-label="Metric" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e96 "><strong class="uicontrol">MCC</strong></td>
            <td data-label="Description" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e98 ">
              <p>The <em>Matthews Correlation Coefficient</em> (MCC)
                is a model quality measure. MCC provides a more even
                representation of the four parts of the confusion
                matrix than other classification
                metrics.</p>
              <p>Range:</p>
              <ul>
                <li>+1 means that the model correctly predicts the
                  class every time.</li>
                <li>0 means that the model is as good as random
                  guessing</li>
                <li>-1 means that the model wrongly predicts the
                  opposite class every time</li>
              </ul>
            </td>
          </tr>
          <tr class="">
            <td data-label="Metric" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e96 "><strong class="uicontrol">Informedness</strong></td>
            <td data-label="Description" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e98 ">
              <p><em>Informedness</em> measures how informed the model
                is about positives and
                negatives.</p>
              <p>Formula:</p>
              <p>Informedness = TPR
                + TNR - 1</p>
              <p>where</p>
              <ul>
                <li>TPR is the True Positive Rate</li>
                <li>TNR is the True Negative Rate</li>
              </ul>
            </td>
          </tr>
          <tr class="">
            <td data-label="Metric" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e96 "><strong class="uicontrol">Accuracy</strong></td>
            <td data-label="Description" class="slds-cell-wrap" style="vertical-align:top;" headers="d163365e98 ">
              <p><em>Accuracy</em> measures the proportion of outcomes
                that the model predicted
                correctly.</p>
              <p>Formula:</p>
              <p>Accuracy = (TP +
                TN) / Total # of Predictions</p>
              <p>where</p>
              <ul>
                <li>TP represents the number of true
                  positives (positive prediction with a positive
                  result)</li>
                <li>TN represents the number of true
                  negatives (negative prediction with a negative
                  result)</li>
                <li>Total # of Predictions that the model
                  made (correct predictions plus incorrect
                  predictions)</li>
              </ul>
            </td>
          </tr>
        </tbody>
      </table>

## Confusion Matrix for All Classes

This chart shows you at a glance how accurately the model predicts each class. Where the row and column for each class intersect in the table, the color of the square indicates the model’s accuracy. The legend shows a color progression from 0 to 1, with 1 representing 100% accuracy. In this example, the darker blue diagonal indicates a high percentage of true positives.

#### See Also

- [Metrics for Multiclass Classification Use Cases](https://help.salesforce.com/s/articleView?id=analytics.bi_edd_model_metrics_multiclass.htm&release=264.0.0&type=5 "The multiclass classification use case is based on categorical variables with 3–10 possible outcomes. The Model Metrics tabs show quality statistics associated with multiclass models.")