> For the complete documentation index, see [llms.txt](https://docs.infraon.io/infraon-help/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.infraon.io/infraon-help/infinity-user-guide/asset/aiops-configuration/adaptive-threshold.md). # Adaptive Threshold It helps detect anomalies in performance metrics by automatically learning patterns and setting dynamic thresholds, instead of relying on fixed, static values. It is part of the Prediction engine and is particularly useful for identifying unusual behavior in real-time monitoring data. Adaptive Thresholds are **model-generated** limits that define what is considered "normal" for each metric over time. Instead of using a fixed value (like 80% for all metrics), the system learns from historical data and sets a **custom threshold for each hour** of each day. This ensures better accuracy in detecting genuine anomalies while reducing noise, especially in environments with varying traffic patterns, workloads, or seasonal behavior. ## **How Adaptive Threshold is Created** **Data Collection** The system regularly collects performance metrics (polled data) (e.g., every 5 minutes). **Model Training** The short-term prediction model is trained using the **last 30 days of historical data**, which is aggregated at an hourly level. This model generates adaptive thresholds for identifying anomalies in real time. **Window Forecasting** The model predicts usage for the **next 2 weeks**, broken down **hour by hour**. **Threshold Generation** For each hour, the system generates the following dynamic thresholds: * **Polled Data** – Actual collected data from the resource. * **Lower Band** – Minimum expected value. * **Upper Band** – Maximum expected value. * **Critical Upper Band** – A stricter upper threshold for flagging critical anomalies. These thresholds help the system determine normal vs. abnormal behavior for each time window. ## **How It Works** Adaptive Thresholds use short-term **time-series prediction models** that learn usage patterns over time. Key concepts include: * **Trend detection –** Identifies regular increase/decrease patterns. * **Hourly profiling –** Learns a "normal" value for each specific hour. * **Upper and lower bounds –** Represent a statistical confidence range. * **Model refresh –** Rebuilt every 7 days using the latest data. The system avoids overfitting or underfitting by using a balanced model sensitivity. ## **What do you see on the screen** The **Adaptive Threshold** screen displays a performance graph that visually compares actual system behavior against dynamically learned threshold ranges. The graph provides a real-time, hour-by-hour overview of how a metric (e.g., CPU, memory, interface utilization) performs relative to predicted safe zones. **Adaptive Threshold |** Basic Details


Label	Action	Description / Example
Polled Data	Displays actual usage	Blue line showing real-time performance values collected at intervals.
Lower Band	Shows the minimum predicted value	Bottom boundary of the threshold range. Values below this are considered anomalies.
Upper Band	Shows the maximum predicted value	Top boundary of the expected range. Values above this may indicate a warning or anomaly.
Critical Upper Band	Flags high-risk upper threshold	Stricter limit above the Upper Band used to detect critical-level anomalies.
Red Markers	Highlights anomalies	Appears when actual data goes outside the threshold bands.
Filters
Current Month	Time filter	Displays data from today to the end of the current month.
Next Month	Time filter	Displays projected threshold behavior for the following month.
Next 3 Months	Time filter	Shows threshold forecasts and anomalies across the next 90 days.
Next 6 Month	Time filter	Displays long-term prediction and anomaly trend for up to 180 days.

**Adaptive Threshold |** Basic Details This provides a detailed view of predicted thresholds and anomalies for each timestamp.


Label	Description / Example
Time	Timestamp for each data point (e.g., May 18, 2025 12:00 AM).
Upper Band	The maximum expected value for the metric at that time (e.g., 8.37%).
Lower Band	The minimum expected value for the time slot (e.g., 6.19%).
Critical Upper Band	A stricter limit highlights critical anomalies (e.g., 9.47%).
Anomaly	Indicates whether the actual value exceeded thresholds (e.g., 20%).
Toggle: Show Only Anomalies	Filters the table to display only rows where anomalies are detected.

{% hint style="info" %} **Export** The adaptive threshold graph can be downloaded as a **PDF**, while the **adaptive threshold Data** table can be exported in **CSV format** for detailed offline review or further processing. {% endhint %} **Note**


Feature	Forecast	Prediction
Focus	Long-term trends (1–6 months)	Short-term anomalies (1–4 weeks)
Use Case	Capacity planning	Anomaly detection
Output	Graphs and capacity reports	Adaptive thresholds and alerts

--- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.infraon.io/infraon-help/infinity-user-guide/asset/aiops-configuration/adaptive-threshold.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.