Amazon Fraud Detector Update: This release adds support for ML Explainability to display model variable importance value in Amazon Fraud Detector.

Author: AWS

.changes/next-release/feature-AmazonFraudDetector-3b76145.json

@@ -0,0 +1,6 @@
+{
+    "type": "feature",
+    "category": "Amazon Fraud Detector",
+    "contributor": "",
+    "description": "This release adds support for ML Explainability to display model variable importance value in Amazon Fraud Detector."
+}

services/frauddetector/src/main/resources/codegen-resources/service-2.json

@@ -71,7 +71,8 @@
       "errors":[
         {"shape":"ValidationException"},
         {"shape":"InternalServerException"},
-        {"shape":"AccessDeniedException"}
+        {"shape":"AccessDeniedException"},
+        {"shape":"ResourceNotFoundException"}
       ],
       "documentation":"<p>Creates a batch prediction job.</p>"
     },
@@ -2248,7 +2249,7 @@
         },
         "eventVariables":{
           "shape":"EventVariableMap",
-          "documentation":"<p>Names of the event type's variables you defined in Amazon Fraud Detector to represent data elements and their corresponding values for the event you are sending for evaluation.</p>"
+          "documentation":"<p>Names of the event type's variables you defined in Amazon Fraud Detector to represent data elements and their corresponding values for the event you are sending for evaluation.</p> <important> <ul> <li> <p>You must provide at least one eventVariable</p> </li> <li> <p>If detectorVersion is associated with a modelVersion, you must provide at least one associated eventVariable</p> </li> </ul> </important> <p>To ensure highest possible fraud prediction and to simplify your data preparation, Amazon Fraud Detector will replace all missing variables or values as follows:</p> <p> <b>For Amazon Fraud Detector trained models:</b> </p> <p>If a null value is provided explicitly for a variable or if a variable is missing, model will replace the null value or the missing variable (no variable name in the eventVariables map) with calculated default mean/medians for numeric variables and with special values for categorical variables.</p> <p> <b>For External models ( for example, imported SageMaker):</b> </p> <p>If a null value is provided explicitly for a variable, the model and rules will use “null” as the value. If a variable is not provided (no variable name in the eventVariables map), model and rules will use the default value that is provided for the variable. </p>"
         },
         "externalModelEndpointDataBlobs":{
           "shape":"ExternalModelEndpointDataBlobMap",
@@ -2633,6 +2634,10 @@
       "type":"string",
       "enum":["DETECTORPL"]
     },
+    "ListOfLogitMetrics":{
+      "type":"list",
+      "member":{"shape":"LogitMetric"}
+    },
     "ListOfModelScores":{
       "type":"list",
       "member":{"shape":"ModelScores"}
@@ -2680,6 +2685,29 @@
         }
       }
     },
+    "LogitMetric":{
+      "type":"structure",
+      "required":[
+        "variableName",
+        "variableType",
+        "variableImportance"
+      ],
+      "members":{
+        "variableName":{
+          "shape":"string",
+          "documentation":"<p>The name of the variable.</p>"
+        },
+        "variableType":{
+          "shape":"string",
+          "documentation":"<p>The type of variable.</p>"
+        },
+        "variableImportance":{
+          "shape":"float",
+          "documentation":"<p>The relative importance of the variable.</p>"
+        }
+      },
+      "documentation":"<p>The logit metric details.</p>"
+    },
     "MetricDataPoint":{
       "type":"structure",
       "members":{
@@ -3386,6 +3414,10 @@
         "trainingMetrics":{
           "shape":"TrainingMetrics",
           "documentation":"<p>The training metric details.</p>"
+        },
+        "variableImportanceMetrics":{
+          "shape":"VariableImportanceMetrics",
+          "documentation":"<p>The variable importance metrics.</p>"
         }
       },
       "documentation":"<p>The training result details.</p>"
@@ -3800,6 +3832,16 @@
       "max":25,
       "min":1
     },
+    "VariableImportanceMetrics":{
+      "type":"structure",
+      "members":{
+        "LogitMetrics":{
+          "shape":"ListOfLogitMetrics",
+          "documentation":"<p>List of variable metrics.</p>"
+        }
+      },
+      "documentation":"<p>The variable importance metrics details.</p>"
+    },
     "VariableList":{
       "type":"list",
       "member":{"shape":"Variable"}