How to cluster discrete data

Hi!

I have a database containing discrete features. For example, number of hairpinloops, number of elements, length of a sequence, the % of A nucleotides. Now I would like to apply some clustering algorithms. Does anyone know which algorithms in matlab are suited for discrete data?

Thanks a lot, Iene

Answer 1

[Purvaja]

Hi @Iene Rutten,

There are various ways to obtain clusters. You can refer the following methods:

• K-Means clustering: The function “k-means" partitions data into k mutually exclusive clusters and returns the index of the cluster to which it assigns each observation. Requires number of clusters. (https://www.mathworks.com/help/stats/k-means-clustering.html )

    [idx, C] = kmeans(data, k); % k is the number of clusters 

• K-medoids Clustering: “K-medoids” is like “K-means” but is more robust to noise and outliers. Requires number of clusters too. (https://www.mathworks.com/help/stats/kmedoids.html)

    [idx, C] = kmedoids(data, k); % k is the number of clusters 

• DBSCAN (Density-Based Spatial Clustering of Applications with Noise): Unlike “k-means” clustering, the ”DBSCAN” algorithm does not require prior knowledge of the number of clusters. It works with distance metrics and can be applied to discrete data.(https://www.mathworks.com/help/stats/dbscan-clustering.html)

    epsilon = 0.5; % Distance threshold 
    minPts = 5; % Minimum number of points to form a cluster 
    idx = dbscan(data, epsilon, minPts); 

• Gaussian Mixture Models (GMM): “GMM” clustering can accommodate clusters that have different sizes and correlation structures within them.(https://www.mathworks.com/help/stats/clustering-using-gaussian-mixture-models.html)

    gm = fitgmdist(data, k); % k is the number of clusters 
    idx = cluster(gm, data); 

To check out more methods, you can refer to the following resource:

https://www.mathworks.com/discovery/clustering.html

You can also access release-specific documentation using these commands in your MATLAB command window:

    web(fullfile(docroot, 'stats/k-means-clustering.html')) 
    web(fullfile(docroot, 'stats/kmedoids.html')) 
    web(fullfile(docroot, 'stats/dbscan-clustering.html')) 
    web(fullfile(docroot, 'stats/clustering-using-gaussian-mixture-models.html')) 

Hope this helps you!