You are performing a filter based feature selection for a dataset 10 build a multi class classifies by using Azure Machine Learning Studio. The dataset contains categorical features that are highly co

Question

Options

• AChi-squared
• BSpearman correlation
• CKendall correlation
• DPerson correlation

• A
  16% (5)
• B
  3% (1)
• C
  6% (2)
• D
  74% (23)

Explanation

NOTE: The stated correct answer (D, Pearson correlation) is questionable from a statistical standpoint. Pearson correlation measures the linear relationship between two continuous (numerical) variables. For categorical features correlated with a categorical multi-class label, Chi-squared (A) is the statistically appropriate method, as it tests independence between categorical variables. Spearman and Kendall are rank-based correlation methods suited for ordinal data. In the context of Azure Machine Learning Studio's Filter Based Feature Selection module, the scoring method must match the data type - if the exam intends that the features or label are treated as continuous, Pearson applies, but for purely categorical features with a categorical label, Chi-squared is the correct choice. Many exam versions flag this as a known answer-key discrepancy.

No community discussion yet for this question.