KMeans Interview Questions

❓ What is the primary goal of KMeans clustering?

Answer:
To partition data into K clusters based on similarity, minimizing intra-cluster variance

❓ How does KMeans determine the initial cluster centers?

Answer:
By randomly selecting K data points as initial centroids

❓ What is the main limitation of KMeans regarding cluster shape?

Answer:
It assumes spherical and equally sized clusters, struggling with non-spherical shapes

❓ How do you choose the optimal number of clusters (K) in KMeans?

Answer:
Using methods like the Elbow Method or Silhouette Score

❓ What is the role of the inertia metric in KMeans?

Answer:
Measures the sum of squared distances from each point to its cluster center

❓ Can KMeans handle categorical data directly?

Answer:
No, it requires numerical data; categorical variables must be encoded

❓ How does KMeans handle outliers?

Answer:
Outliers can distort cluster centers and increase inertia

❓ What is the difference between KMeans and KMedoids?

Answer:
KMeans uses mean of points, while KMedoids uses actual data points as centers

❓ Why is feature scaling important for KMeans?

Answer:
To ensure all features contribute equally and prevent dominance by large-scale features

❓ How does KMeans work in high-dimensional spaces?

Answer:
It suffers from the curse of dimensionality, making distance measures less meaningful

❓ What is the time complexity of KMeans?

Answer:
O(n * k * t), where n is samples, k is clusters, and t is iterations

❓ What is the space complexity of KMeans?

Answer:
O(k * d), where k is clusters and d is features

❓ How do you evaluate the quality of KMeans clustering?

Answer:
Using metrics like silhouette score, within-cluster sum of squares, or Davies-Bouldin index

❓ Can KMeans be used for image segmentation?

Answer:
Yes, by treating pixel values as features and clustering them

❓ How does KMeans initialize centroids differently in KMeans++?

Answer:
Centroids are initialized to be far apart, improving convergence speed and quality

❓ What happens if the number of clusters (K) is too small?

Answer:
Clusters may be overly broad, merging distinct groups

❓ What happens if the number of clusters (K) is too large?

Answer:
Overfitting occurs, creating artificial clusters

❓ Does KMeans guarantee a global optimum?

Answer:
No, it converges to a local optimum depending on initialization

❓ How can you improve KMeans performance on large datasets?

Answer:
Using MiniBatchKMeans or sampling techniques

❓ What is the effect of random seed on KMeans results?

Answer:
Different seeds lead to different initial centroids, affecting final clusters

#️⃣ #kmeans #machine_learning #clustering #data_science #ai #python #coding #dev

By: t.iss.one/DataScienceQ 🚀