clusteval

• Fix for scatter. Update in input parameters

• Fixes for tSNE perplexity and title
• add normalization parameter
• Various other improvements
• Docstrings update

• many fixes and updates
• wget is replaces by internal functions
• plots are updated
• added new plot: ce.plot_silhouette()
• bugfix for min_samples fixed
• jitter functionality added
• logging improved
• update docstrings

• fix in case silscore is empty

• sklearn changed into scikit-learn in both requirements and setup file.

• Parameter savefig can now be used to store figures in plot, scatter and dendrogram

Example:

from clusteval import clusteval
X, y = make_blobs(n_samples=600, centers=[[1, 1], [-1, -1], [1, -1]], cluster_std=0.4, random_state=0)
ce = clusteval(cluster='agglomerative', evaluate='derivative')
results = ce.fit(X)

ce.plot(savefig={'fname':'clusteval_plot.png'})
ce.scatter(X, dot_size=100, savefig={'fname':'clusteval_scatter.png'})
ce.dendrogram(savefig={'fname':'clusteval_dendrogram.png'})

• added DOI

• Input parameter "method" is changed into "evaluate"
• Added docstrings
• Some code refactoring and cleaning

• improved default parameters for dbscan to detect clusters of various densities

• bug fix in the special case when no clusters are possible to detect

• check for correctness of the input parameters cluster and method.
• better error handling for hdbscan

• bugfix for hdbscan: AttributeError: 'clusteval' object has no attribute 'results'
• Added licence in setup.py

• fix for dendrogram cut in special cases

• bug Fix in case no clusters are detected

Huge structural update.
The input parameter "method" can only be used for the evaluation of methods, such as silhouette, dbindex and derivative. The clustering approach can separately be chosen using the input parameter "cluster", which can be: agglomerative, kmeans, dbscan, hdbscan.

• added input parameter: "cluster"
• input parameter "method" is for: silhouette, dbindex, derivative
• Changed input parameter minclusters to min_clust
• Changed input parameter maxclusters to max_clust
• Pep8 styling
• code refactoring
• Docstrings updated

• dendrogram plotting possible
• Docstring updates

• Fix for hdbscan plotting

Huge and happy update in clusteval

• Code refactored into classes
• docstrings updated
• plot centralized
• scatter centralized
• code cleaning
• examples added

Have fun!