MLS-C01 Question 68
Single answerYou are working with a dataset of customer transactions for a retail company. Before building a machine learning model to predict customer churn, you want to perform exploratory data analysis (EDA) to better understand the dataset. During EDA, you notice that one of the features, 'Annual Income', has a skewed distribution with outliers on the higher end. What is the MOST appropriate step to address this during EDA?
- A
Apply a logarithmic transformation to the 'Annual Income' feature
- B
Remove all rows with 'Annual Income' values greater than one standard deviation from the mean
- C
Normalize the 'Annual Income' feature using min-max scaling
- D
Leave the 'Annual Income' feature as is, since machine learning models can handle skewed data
Show answer and explanation
Correct answer: A
Explanation
When performing exploratory data analysis, addressing skewed data is critical to ensure the machine learning model performs well. A logarithmic transformation is a standard technique used to reduce skewness and make the distribution more normal. This step helps improve the model's ability to learn from the data, particularly for algorithms sensitive to feature distributions.
- A. Correct.
Applying a logarithmic transformation is an effective way to handle skewed data. It reduces the impact of extreme values and makes the distribution more normal, which can improve model performance.
- B. Incorrect.
Removing rows based on an arbitrary threshold, such as one standard deviation, can lead to significant data loss and bias, especially if the feature contains important information.
- C. Incorrect.
While normalization (e.g., min-max scaling) is useful for scaling data, it does not address the skewness issue. Skewed data will remain skewed even after normalization.
- D. Incorrect.
Leaving the feature as is may work in some cases, but highly skewed data can negatively impact the performance of certain machine learning algorithms, especially those that assume normally distributed features.