Consider the following points if the functions display the accumulated columns as the first columns of the output:
- The data type of the first column cannot be BLOB or CLOB.
- The first column becomes the Primary Index and must be selected carefully.
- The Primary Index column (by default, the first column) affects the data distribution and performance. The column with more unique values must be the Primary Index column and vice versa. If the first column in the output is the accumulated column, then you must select the Primary Index column explicitly (if the default column is not optimal) for better performance.
The following functions add the accumulated columns at the end of the output:
- NaiveBayesPredict
- DecisionTreePredict
- Pack
- Unpack
- TD_KMeansPredict
- TD_Silhouette
The following functions add the accumulated columns at the beginning of the output:
- GLMPredict
- DecisionForestPredict
- StringSimilarity
- NgramSplitter
- TD_GetRowsWithMissingValues
- TD_GetRowsWithoutMissingValues
- TD_ConvertTo
- TD_qqnorm
- TD_TextParser
- TD_NumApply
- TD_StrApply
- TD_RoundColumns
- TD_BincodeTransform
- TD_NonLinearCombineTransform
- TD_OrdinalEncodingTransform
- TD_PolynomialFeaturesTransform
- TD_RowNormalizeTransform
- TD_ScaleTransform
- TD_RandomProjectionTransform
- TD_SentimentExtractor