Outlier Detection Using Percentile-Based Method
This guide explains how to create a Sentinel to detect salary outliers using percentile-based detection (e.g., top 1% or bottom 1% of salaries).
How Percentile-Based Detection Works
This method identifies outliers as values that fall above a high percentile (e.g., top 1%) or below a low percentile (e.g., bottom 1%) of the data distribution.
Example Dataset
Download the sample dataset as a CSV file: outlier-dataset.csv
Example Alert Condition
In the Alert Condition field, enter:
Find employees whose salary is in the top 1% or bottom 1% of all salaries.
This will flag any records where the salary is unusually high or low compared to the rest of the dataset.
Example Alert Response
When outliers are detected, the alert validation will show a response like this:
Alert Condition: Find employees whose salary is in the top 1% or bottom 1% of all salaries.
Alert Triggered! Your alert condition would trigger based on the sample data.
Execution Details:
- Alert Condition Evaluation: True (Alert would trigger)
- Code Execution Status: Successful
Matching Records (2 rows):
| department | emp_id | joining_date | name | salary |
|---|---|---|---|---|
| Engineering | 18 | 2023-01-10T00:00:00 | Rachel Allen | 83000 |
| Engineering | 21 | 2025-05-22T00:00:00 | Jhonny English | 0 |
This means the sentinel has detected employees whose salaries are in the extreme percentiles of the dataset.
Tips
- You can change the percentage in the alert condition (e.g., top 5%, bottom 5%) to adjust sensitivity.
- Use the validation feature to preview which records will be flagged as outliers.