# Review inference results
<a name="reviewing-inference-results"></a>

## Retrieve inference results
<a name="retrieving-inference-results"></a>

### Latest inference results
<a name="get-latest-inference-results"></a>

Run the following command to fetch the most recent inference result for an asset property. For more information, see [ get-asset-property-value](https://docs.aws.amazon.com/cli/latest/reference/iotsitewise/get-asset-property-value.html) in the *AWS CLI Command Reference Guide*.

```
aws iotsitewise get-asset-property-value \
    —asset-id asset-id \
    —property-id result-property-id
```

### Inference results history
<a name="get-inference-results-history"></a>

Run the following command to fetch the history of inference results for a specified time window. For more information, see [ get-asset-property-value-history](https://docs.aws.amazon.com/cli/latest/reference/iotsitewise/get-asset-property-value-history.html) in the *AWS CLI Command Reference Guide*.

```
aws iotsitewise get-asset-property-value-history \
    —asset-id asset-id \
    —property-id result-property-id \
    —start-date start-time \
    —end-date end-time
```

### Example response
<a name="example-response"></a>

**Example of an inference result response:**  

```
{
  "value": {
    "stringValue": "{\"timestamp\": \"2025-02-10T22:42:00.000000\", \"prediction\": 0, \"prediction_reason\": \"NO_ANOMALY_DETECTED\", \"diagnostics\": [{\"name\": \"asset-id\\\\property-id\", \"value\": 0.53528}]}"
  },
  "timestamp": {
    "timeInSeconds": 1739227320,
    "offsetInNanos": 0
  },
  "quality": "GOOD"
}
```

### Response fields
<a name="response-fields"></a>
+ **value.stringValue** – A JSON string containing the inference result with the following fields:
  + **timestamp** – The timestamp of the TQV against which inference is performed.
  + **prediction** – The prediction result (0 for no anomaly, 1 for anomaly detected).
  + **prediction\$1reason** – The reason for the prediction (`NO_ANOMALY_DETECTED` or `ANOMALY_DETECTED`).
  + **diagnostics** – An array of diagnostic information showing contributing factors.
+ **timestamp** – The timestamp when the result is recorded in AWS IoT SiteWise.
+ **quality** – The quality of the data point (typically `GOOD`).

## Understand inference results
<a name="understanding-inference-results"></a>

An inference result returned by AWS IoT SiteWise anomaly detection includes key information about the model's prediction at a specific timestamp, including whether an anomaly was detected and which sensors contributed to the anomaly.

**Example of a detailed inference result:**  

```
{
    "timestamp": "2021-03-11T22:25:00.000000",
    "prediction": 1,
    "prediction_reason": "ANOMALY_DETECTED",
    "anomaly_score": 0.72385,
    "diagnostics": [
      { "name": "asset_id_1\\\\property_id_1", "value": 0.02346 },
      { "name": "asset_id_2\\\\property_id_2", "value": 0.10011 },
      { "name": "asset_id_3\\\\property_id_3", "value": 0.11162 }
    ]
}
```

The `diagnostics` field is useful for interpreting *why* the model makes a certain prediction. Each entry in the list includes:
+ `name`: The sensor that contributed to the prediction (format: `asset_id\\\\property_id`).
+ `value`: A floating-point number between 0 and 1, representing the **relative weight or importance**, of that sensor at that point in time.

User benefits:
+ Understand which sensors had the strongest impact on an anomaly.
+ Correlate high-weight sensors with physical equipment behavior.
+ Inform root cause analysis.

**Note**  
Even when `prediction = 0` (normal behavior), the diagnostics list is returned. This helps assess which sensors are currently influencing the model's decisions, even in healthy states.