In CRISP-DM, the data preparation step includes reducing the dataset by removing irrelevant attributes or records.

• A. Data Understanding
• B. Data Preparation
• C. Modeling
• D. Evaluation

Answer: (B)

In the data preparation stage, the focus is on getting the data ready for modeling by cleaning, transforming, and shaping it. Reducing the dataset by removing attributes that don’t help the model (irrelevant features) and by filtering out unnecessary or noisy records is a core part of this stage. This pruning, which includes feature selection and sampling, lowers dimensionality and data noise, speeds up training, and often improves model generalization by keeping only information that is actually useful for the task at hand.

Other phases have different roles: data understanding is about exploring and assessing the data’s quality and characteristics, modeling is about choosing algorithms and building predictive models, and evaluation is about judging how well those models meet the objectives. The act of reducing the dataset through removing irrelevant attributes or records belongs to preparing the data for modeling, not to understanding, modeling itself, or evaluation.