K-Medians Clustering for Beginners

K-medians is a machine learning algorithm used to group data into clusters, especially useful when your data includes outliers or non-normal distributions.

Unlike K-means (which uses the average), K-medians uses the median value as the center of each cluster — making it more stable and less sensitive to unusual or extreme data points.

1. What Is Clustering?
Clustering is a type of unsupervised learning used to automatically group data points that are similar. It's often used for:
- Pattern recognition
- Customer segmentation
- Anomaly detection
- Recommendation systems
K-medians is especially useful when mean values can be misleading.

2. K-Means vs. K-Medians

Feature	K-Means	K-Medians
Center type	Mean (average)	Median (middle value)
Outlier handling	Sensitive to extreme values	Robust to outliers
Speed	Faster (uses squared distances)	Slower (uses Manhattan distance)
Distance metric	Euclidean	Manhattan (L1)

3. How K-Medians Works
- Choose K (the number of clusters you want)
- Pick random medians as initial centers
- Assign each data point to the nearest median (using Manhattan distance)
- Recalculate the median of each cluster
- Repeat steps 3–4 until clusters stabilize (converge)

4. When to Use K-Medians
- Your data has outliers or skewed values
- You need robust clustering
- You're working with 1D or low-dimensional data
- You want a simpler model than density-based clustering

5. Example Use Case
Suppose you're analyzing transaction amounts from an e-commerce site. Most purchases are normal, but a few are either errors or fraud attempts. Using K-means might pull cluster centers toward the extreme values.
K-medians would give you more representative groups by ignoring those outliers when calculating the center.

6. Limitations
- Slower than K-means on large datasets
- Harder to compute medians in high-dimensional space
- Less popular in standard libraries (but still supported in tools like scikit-learn-contrib)

7. Code Snippet (Python-style Pseudocode)
from sklearn_extra.cluster import KMedians
model = KMedians(n_clusters=3, metric='manhattan')
model.fit(data)
labels = model.labels_


Summary
K-medians is a reliable, robust clustering method when you want clean groups and your data isn’t perfectly clean. It’s especially useful in cybersecurity, finance, or real-world sensor data — any scenario where anomalies matter and means lie.