Which of the following is MOST important to have in place when initially populating data into a data frame for an AI model?

Question

Options

• AThe box charts, histograms, scatterplots, and Venn diagrams that identify correlations and outliers
• BThe code for separating data into training and testing data sets
• CAn analysis of exploratory data that checks for incorrect data types, null values, and duplicate
• DAn approved risk assessment for including, excluding, or subsequently dropping data attributes

• B
  4% (1)
• C
  92% (22)
• D
  4% (1)

Explanation

Exploratory data analysis (EDA) that checks for incorrect data types, null values, and duplicate records is the foundational first step when populating a data frame. These data quality issues - if undetected - corrupt every subsequent stage of the ML pipeline: biased training, incorrect splitting, and ultimately unreliable predictions. Identifying and resolving them at the outset is essential. Visualizations (A) are a valuable EDA output but come after basic quality checks are in place. Train/test splitting code (B) is a later pipeline step that only makes sense after the data is validated and clean. A risk assessment for data attributes (D) is an important governance activity but is a prerequisite decision made before data collection, not the first technical step when populating the data frame.

No community discussion yet for this question.