Which preprocessing concern is highlighted as important for improving k-means results?

• A. Higher Dimensionality Always Improves Clustering
• B. Missing Values Should Be Ignored by K-Means
• C. Categorical Variables Do Not Require Encoding
• D. Outlier Values Can Distort the Mean and Should Be Addressed

Answer: (D)

In k-means clustering, the centers (centroids) are computed as the mean of the points assigned to each cluster. The mean is sensitive to extreme values, so outliers can pull a centroid away from where most of the data actually lie. This shifts the cluster boundaries, causes misassignments, and leads to less cohesive clusters overall. Addressing outlier values before running k-means—through removal, capping, or robust alternatives—helps the centroids reflect the true structure of the data and produces more accurate, stable clustering.

It's also worth noting that scaling features is important for distance-based methods like k-means, and missing values or categorical features require appropriate handling or encoding, but the preprocessing concern highlighted here is that outliers can distort the mean and should be addressed to improve results.