HomeTemplatesAnomaly & Outlier Detection
Data Quality & EDA

Anomaly & Outlier Detection

Detect statistical anomalies and distribution distortions in your data streams.

Recipe Overview & Methodology

Identify abnormal events and measurement spikes before they corrupt downstream reports. Uses statistical IQR boundaries, rolling z-score deviations, and distribution kurtosis scans.

Statistical Validation Rules

  • Tukey's Fences method with 1.5x and 3.0x Interquartile Range (IQR).
  • Modified Z-Score using Median Absolute Deviation (MAD) for robust scaling.
  • Temporal moving window anomaly filtering for time-series metrics.
  • Automated alert trigger payloads for workflow pipelines.

Sample Output Metrics

Outlier Rate
1.2%
14 / 1,200 rows
Max Z-Score
+4.61
Extreme positive anomaly
Distribution Skew
1.84
Right-skewed tail
Recommended Action
Winsorize / Flag
Clean before modeling
Execution LogicClustey Engine
# Statistical Outlier Boundary Scan
q25, q75 = dataset['metric'].quantile([0.25, 0.75])
iqr = q75 - q25
lower_bound = q25 - (1.5 * iqr)
upper_bound = q75 + (1.5 * iqr)

anomalies = dataset[(dataset['metric'] < lower_bound) | (dataset['metric'] > upper_bound)]
Ready to apply this recipe to your own dataset?
Connect your dataset and generate this analysis in one click.
Try this in demo