TD_OutlierFilterTransform
TD_OutlierFilterTransform is designed as a cleanup step after outlier thresholds have already been learned by TD_OutlierFilterFit. In practice, the fit step calculates the normal operating range for each target column using the selected method, such as percentiles, and stores those lower and upper bounds in a fit table.
TD_OutlierFilterTransform then reads those learned bounds and applies them to new input data, removing any rows whose values fall outside the accepted range. In other words, it acts as a filter: rows that are considered anomalous are excluded from the transformed output, while rows that remain within the learned limits are kept.
This makes it especially useful for preprocessing data before modeling, reporting, or downstream analysis, where you want to retain only the “normal” portion of the data and suppress extreme spikes or dips.
Teradata is a strong fit for this because it handles large-volume, set-based data cleaning in-database, so outlier filtering can be applied efficiently at scale without moving data around. It is especially useful when you want to remove anomalous rows before reporting, aggregation, or model training while keeping the process fast, repeatable, and SQL-native.
Below is a screen shot of the demo i created and a link to download the script
