Time Series Anomaly Detection
The Data Service provides a Time Series Anomaly Detection algorithm that uses intelligent time-series analysis to accurately detect real-time anomalies in your device data.
Basic information
On the Instance Overview page of the IoT Platform console, click the target Enterprise instance.
In the navigation pane on the left, choose .
-
In the algorithm template list, find the Time Series Anomaly Detection algorithm and click View in the Actions column.
-
On the Basic Information tab, you can view information such as the algorithm's features and use cases.
Instructions
-
In the algorithm template list, find the Time Series Anomaly Detection algorithm and click View in the Actions column.
-
On the Instructions tab, click Input Parameter Format and Output Parameter Format to view the parameter descriptions.
-
Input data: The data required to run the Time Series Anomaly Detection algorithm.
-
Output data: The data generated by the Time Series Anomaly Detection algorithm.
The input data format (anomaly detection data table) includes the parameters
val(DOUBLE, the data value to analyze) andts(BIGINT, the data's Unix timestamp). The output data format (anomaly detection result table) includes these parameters:val(DOUBLE, the original data value),baseline_val(DOUBLE, the predicted normal baseline value; values exceeding this are outliers),ts(BIGINT, the data's Unix timestamp), andis_outlier(BOOLEAN, whether the data point is an outlier). -