Intelligent baseline

Updated at:

Select device operation metrics at the product level for anomaly detection, and enable alert notifications to reduce static threshold alerts and false alarms.

Prerequisites

Background information

Before you use the intelligent baseline feature, understand the following concepts:

  • Intelligent baseline training algorithm: Runs daily in the early morning to generate a prediction model for each intelligent baseline.

  • Intelligent baseline prediction algorithm: Runs every minute to produce a metric value based on the current metric data and the prediction model.

    image

Create intelligent baseline

  1. Log on to the IoT Platform console.

  2. In the upper-left corner of the console, select the region where your target Enterprise Edition instance is located. Then, on the Instance Overview page, find the corresponding Enterprise Edition instance and single click its name.

  3. In the left-side navigation pane, select Monitoring & Operations >Device Diagnostics, and then single click Intelligent Baseline.

  4. On the Intelligent Baseline page, single click Create Intelligent Baseline. In the dialog box that appears, configure the parameters and then single click OK.

    Parameter

    Description

    Intelligent Baseline Name

    Assign a name to the intelligent baseline for easy identification. It supports Chinese, English letters, numbers, and underscores (_), up to 32 characters in length. You can modify the name after creation.

    Once the intelligent baseline is established, it supports modifications.

    Chart

    Choose the detection metric for the intelligent baseline. For available metrics, refer to Metric Description in this topic.

    Product

    Select the product to be monitored by the intelligent baseline, either All Products or a specific product.

    Alert Sensitivity

    The alert sensitivity level determines the frequency of alert generation. You can adjust this setting after creating the intelligent baseline.

    Once the intelligent baseline has been established, it supports modifications.

View intelligent baseline detection information

After the intelligent baseline is created, you can view its details in the list.

The Status column shows the daily training status of the intelligent baseline.

  • Training: The intelligent baseline has been created for less than 14 days and training is not yet complete.

  • Training Abnormal: The intelligent baseline has existed for at least 14 days, but training failed due to insufficient metric data or average metric values below the minimum training threshold.

    Note

    The training algorithm updates the status after each early-morning run. If the status shows Training Abnormal, check again the next day.

  • Training Successful: The intelligent baseline has existed for at least 14 days, and the metric data volume and average values meet the training requirements.

    Note

    After successful training, the prediction model remains valid for 30 days. Alerts are generated normally during this period. If training fails for 30 consecutive days, the model becomes invalid.

  • Single click the icon

    The metric line chart displays:

    Chart Data Item

    Description

    Metric Value

    Shows the real-time value of the metric.

    Intelligent Baseline

    The baseline value predicted by the machine learning algorithm based on historical data.

    Anomaly

    Whether the current metric value is abnormal, as determined by the prediction algorithm.

    Important

    Common reasons why intelligent baseline values may be absent from the chart:

    • The intelligent baseline has existed for less than 14 days and model training has not started.

    • The intelligent baseline has existed for at least 14 days, but training has not succeeded due to insufficient metric data or average values below the minimum threshold.

    • The intelligent baseline was successfully trained, but more than 30 days have passed since the last successful training, so the model is no longer valid.

Enable alert notification push

In the intelligent baseline list, single click the Alert Notification Push column's

Before using intelligent baseline alerts for the first time, click Alert Notification Configuration at the top of the page. In the Alert Notification Configuration panel, select the alert notification recipient and alert level, and click OK.

After configuring alert notifications, click View Alert Records in the Operations column to review and manage alerts in batch.

Metric description

The intelligent baseline feature supports the following metrics:

Metric Item

Data Metric

Messaging

Message volume sent from the platform via the MQTT protocol

Message volume sent from the platform (HTTP/2)

Message volume received by the platform via HTTP/2

Message volume received by the platform via MQTT

Number of messages sent from the platform to devices

Number of messages sent from devices to the platform

Total message volume for messaging

Number of devices connected via the MQTT protocol

Message volume sent from the platform via the AMQP protocol

Thing Specification Language messages

Number of events sent from devices to the platform

Number of property set operations performed by the platform on devices

Number of service invocation operations performed by the platform on devices

Number of property report operations performed by devices

Number of warning events sent from devices to the platform

Number of fault events sent from devices to the platform

Message forwarding

Number of message forwarding operations to TSDB cloud products by the rules engine

Number of message forwarding operations to REPUBLISH cloud products by the rules engine

Number of message forwarding operations to RDS cloud products by the rules engine

Number of message forwarding operations to OTS cloud products by the rules engine

Number of message forwarding operations to MQ cloud products by the rules engine

Count of message forwards by the rules engine to the Simple Message Queue (formerly MNS) service.

Number of message forwarding operations to FC cloud products by the rules engine

Number of message forwarding operations to DATAHUB cloud products by the rules engine

Delay in message forwarding by the rules engine

Total number of message forwarding operations by the rules engine

Message size in rules engine forwarding