Databricks Machine Learning Associate Question 141
Select 3You are working on a machine learning project in Databricks and have loaded a dataset containing customer transaction data. Your goal is to perform exploratory data analysis (EDA) to understand the distribution of numeric features and detect potential data quality issues. Which of the following steps should you prioritize during EDA?
- A
Visualize the distribution of numeric features using histograms or boxplots.
- B
Generate summary statistics (e.g., mean, median, standard deviation) for numeric columns.
- C
Immediately train a baseline machine learning model to understand feature importance.
- D
Check for missing values and assess their distribution across the dataset.
- E
Perform hyperparameter tuning to optimize model performance.
Show answer and explanation
Correct answers: A, B, D
Explanation
Exploratory Data Analysis (EDA) focuses on understanding the structure, distribution, and quality of data. Steps such as visualizing numeric features, generating summary statistics, and identifying missing values are critical for uncovering patterns and potential issues in the dataset. Actions like training models or hyperparameter tuning are outside the scope of EDA and should be performed in later stages of the machine learning workflow.
- A. Correct.
Visualizing the distribution of numeric features using histograms or boxplots is a key step in EDA. It helps you understand the spread, outliers, and overall pattern of the data.
- B. Correct.
Generating summary statistics allows you to capture key metrics such as central tendency and variability, which are essential to understanding the data.
- C. Incorrect.
Training a baseline machine learning model is not part of EDA. This step comes after the data is well-understood and preprocessed.
- D. Correct.
Checking for missing values is crucial in EDA as missing data can impact model performance and needs to be addressed during preprocessing.
- E. Incorrect.
Hyperparameter tuning is a step in model optimization and is not part of the exploratory data analysis process.