Example Gallery
This gallery contains runnable Python examples showcasing the features,
clustering capabilities, and scikit-learn integrations of scikit-clarans.
Cluster Geometries
Examples demonstrating clustering behavior across various geometric shapes.
Distance Metrics and Representations
Examples demonstrating non-Euclidean metrics, sparse matrices, precomputed graphs, text sequences, time series, and mixed tabular data.
Distance Metrics and Metric Parameters in CLARANS Clustering
Graph Clustering with Precomputed Shortest-Path Distances
Custom Callable Distance Metric: Geospatial Haversine Clustering
Clustering on Asymmetric Precomputed Distance Matrices (D != D^T)
Clustering Incomplete Data with Missing Values (NaN)
Clustering Non-Vector Data: Text Sequences with Edit Distance
Clustering Time Series with Dynamic Time Warping (DTW)
Clustering Mixed-Type Tabular Data with Gower’s Dissimilarity
Seeding and Validation
Examples comparing seeding strategies and evaluating cluster count selection.
User-Provided Initial Candidate Centers for CLARANS
Comparison of Initial Medoid Seeding Strategies in scikit-clarans
Cluster Quality vs. Number of Clusters (Silhouette Analysis)
Inference and Pipelines
Examples showing out-of-sample prediction, distance transformations, and scikit-learn pipeline integration.
Hyperparameter Tuning with GridSearchCV and Pipelines
Fast Precomputed Inference: Compact (n_test, k) vs. Full (n_test, n_train)
Robustness and Anomaly Detection
Examples demonstrating resistance to extreme outliers and distance-based anomaly scoring.
Outlier Robustness: CLARANS (k-Medoids) vs. K-Means
Unsupervised Anomaly Detection via Medoid Distance Scoring
Benchmarks and Convergence
Empirical benchmarks comparing runtime, swap evaluations, memory footprint, and scaling behavior.
Benchmarking CLARANS vs. FastCLARANS Convergence Speed
Side-by-Side Comparison: CLARANS, FastCLARANS, and K-Means
Parameter Sensitivity: num_local vs. max_neighbors
Monitoring Search Progress and Restart Convergence Trajectories
Comparing Cost Evaluation Strategies: Delta vs. Brute-Force