Advanced filters provide a powerful way to refine anomaly detection by focusing on specific data patterns and thresholds. This guide explains how to use these filters effectively, covering their key components and functionality
1. Setting the Alert Sensitivity
Sensitivity controls how anomalies are detected, with three preset levels:
- Low Sensitivity:
- Detects even small changes in the data.
- Use this for scenarios where subtle deviations matter.
- Medium Sensitivity:
- Balances detection, flagging moderate changes without being too lenient or strict.
- Best for general-purpose anomaly tracking.
- High Sensitivity:
- Flags only significant changes in data.
- Ideal for focusing on major deviations while ignoring minor fluctuations.
Choose the sensitivity level that aligns with the importance and frequency of changes in your dataset.
2. Volume Threshold
The volume threshold sets the minimum data volume required for an alert to be triggered.
-
What It Does:
- Ensures that anomalies are only flagged if they occur in a dataset meeting or exceeding the specified size.
- Prevents alerts for insignificant, small-scale events.
-
How to Adjust It:
- A low volume threshold (e.g., 1 or 2) triggers alerts even for small datasets.
- A high volume threshold limits alerts to large datasets, focusing on more impactful anomalies.
3. Adding Filters
Filters allow you to narrow down the data considered for anomaly detection. You can create and apply filters to focus on specific metrics, dimensions, or conditions.
How to Add Filters:
- Go to the Advanced Filters section.
- Click Add Filter and define the filtering criteria.
- For example, filter by region, product category, or time range.
- Save and apply the filters.
4. Importing Filters
If you have predefined filters saved on the dashboard, you can import them for use within the settings.
How to Import Filters:
- Go to the Advanced Filters section.
- Click the Import Filters button.
- Select the filters you want.
Results and Examples
Once you’ve set your filters and thresholds, you’ll start seeing alerts based on the conditions you’ve defined. Here are some simple examples to illustrate:
-
Example 1: Low Sensitivity, Volume Threshold = 3
With low sensitivity, the system will detect even minor deviations in the data, and only datasets with at least 3 entries will trigger alerts. This means if a small change happens in a dataset with 3 or more tickets, it will be flagged as an anomaly. -
Example 2: High Sensitivity, Z-Score Threshold = 5, Volume Threshold = 10
Here, only major changes (with a Z-Score of 5 or higher) in datasets with at least 10 entries will trigger alerts. For instance, a large spike in customer complaints or sales in a large dataset could trigger a critical alert, while smaller fluctuations will be ignored. -
Example 3: Filter by Region, Medium Sensitivity
If you filter by Region = North America, the system will only flag anomalies related to data in this region. Combined with medium sensitivity, it will flag moderate changes, allowing you to focus on anomalies in North America without being overwhelmed by alerts from other regions.
By using these filters effectively, you can get a more focused view of the anomalies that matter most to your business, saving time and making your monitoring more efficient.