Hi,
Is there any method to find the optimal features for cluster analysis using MATLAB? Can we use silhouette plot (cluster width and silhouette coefficient ) to select the most relevant features?\
Thanks in advance
There are several algorithms or techniques via one may analyze data clustering:
https://www.mathworks.com/help/stats/cluster-analysis.html
You may learn these 8 algorithms by coding in Matlab as a fun.
- Hierarchical Clustering Produce nested sets of clusters
- k-Means and k-Medoids Clustering Cluster by minimizing mean or medoid distance, and calculate Mahalanobis distance
- Density-Based Spatial Clustering of Applications with Noise Find clusters and outliers by using the DBSCAN algorithm
- Spectral Clustering Find clusters by using graph-based algorithm
- Gaussian Mixture Models Cluster based on Gaussian mixture models using the Expectation-Maximization algorithm
- Nearest Neighbors Find nearest neighbors using exhaustive search or Kd-tree search
- Hidden Markov Models Markov models for data generation
- Cluster Visualization and Evaluation Plot clusters of data and evaluate optimal number of cluster
Usually calculated as ratio of similarity within a cluster and similarity between clusters i.e. minimum inter-cluster distance / maximum intra-cluster distance, where inter-cluster distance may be measured as distance between centroids of two clusters and intra-cluster distance may be measured as maximum distance between two elements within a cluster.
- Badges
- Science topic