A data dictionary (A) is the authoritative source for understanding:
* Data types and numeric formats
* Valid ranges and interpretations
* Field definitions and business meaning
* Normalization and scaling expectations
Before balancing or preprocessing data, developers must verify that they understand each feature correctly.
The AAIA framework emphasizes that misinterpretation of numeric variables often leads to:
* Incorrect normalization
* Faulty scaling
* Skewed class balancing
* Inaccurate model training
Statistical summaries (C) help identify distributions but cannot validate semantic meaning. Confusion matrices (D) are used after training. Libraries (B) are tools, not sources of interpretation.
References:
AAIA Domain 2: Data Management - Data Dictionaries, Metadata, Data Understanding