AI0-001 Implementing AI Solutions Practice Question

Answer analysis

Option-by-option breakdown

For each option: why learners choose it and why it is or isn't the right answer here.

• ✓

  Cleaning and normalization

  Why this is correct

  Cleaning imputes or removes missing values so the classifier receives complete records, while normalisation rescales features onto a comparable range, preventing large-magnitude features from dominating distance or gradient calculations. Both address the dataset's stated defects.

• ✗

  Outlier removal and binning

  Why it's wrong here

  Outlier removal discards extreme records and binning discretises continuous values; neither fills missing entries nor normalises differing scales. They are tempting because both handle noisy or skewed distributions, which suits datasets dominated by measurement errors or non-linear relationships rather than nulls and magnitude gaps.

• ✗

  Feature selection and dimensionality reduction

  Why it's wrong here

  Feature selection and dimensionality reduction remove or combine columns; they leave missing values unfilled and do not equalise feature magnitudes. They are tempting because both reduce overfitting and training cost, which is valuable when the dataset has many redundant or correlated features rather than gaps and scale differences.

• ✗

  Data augmentation and one-hot encoding

  Why it's wrong here

  Augmentation synthesises new samples and one-hot encoding converts categorical labels into binary columns; neither imputes missing values nor rescales numeric features. They are tempting because both are standard preprocessing steps, but they address class imbalance and categorical encoding respectively.