The Analytics Power Hour #269: The Ins and Outs of Outliers with Brett Kennedy
35 snips
Apr 15, 2025 
Brett Kennedy, a freelance data scientist and author of 'Outlier Detection in Python,' delves into the nuances of outlier detection methods. He compares identifying outliers to obscenity, noting the challenges of definition and detection. The discussion spans techniques such as z-scores and the Median Absolute Deviation, emphasizing the importance of context in data analysis. Kennedy also highlights the human touch needed in distinguishing significant anomalies from normal variations, showcasing the interplay between technology and human insight in deciphering data.
AI Snips
Chapters
Books
Transcript
Episode notes
Not All Outliers Are Problems
- Outliers aren't always problems; in finance, science or marketing, anomalies can reveal opportunities or new discoveries.
- Outlier detection helps identify rare phenomena that merit further investigation.
Use Forecasting for Time Series
- Forecasting methods, including ARIMA and exponential smoothing, offer interpretable ways to detect time-series outliers.
- Combine trend and seasonal patterns to understand and interpret forecast deviations as outliers.
Frame Reference in Outlier Detection
- Use multiple frames of reference to define 'normal' when detecting outliers in time series data.
- Comparing recent values to similar past periods helps contextualize anomalies accurately.
