AutoClusterML 0.1.2

This is a python package useful for the automated dimensionality reduction and clustering (both unsupervised and supervised).

pip install AutoClusterML

from AutoClusterML import AutoCluster

a=AutoCluster.autocluster(data,labels=None,clusters=None)

AutoClusterML

This is a python module for automated dimensionality reduction and clustering (both unsupervised and supervised).

Features

• Dimensionality reduction techniques applied based on user request for method and number of components.

• Automated implementation of clustering algorithms on the data (with and without the presence of labels).

• Scatter plots for all clustering algorithms

• Metric based bar plot comparison of clustering algorithms

• Contingency matrices for all clustering algorithms

• Confusion matrices for all clustering algorithms

dim_reducer

x_p=dim_reducer(x,method='PCA',components=2)

This function automatically reduces the dimension of the data using 10 different dimensionality reduction techniques namely, PCA, factor

analysis, ICA, Incremental PCA, Kernel PCA, Mini Batch Sparse PCA, NMF, Sparse PCA, Mini batch NMF and SVD respectively.

autocluster

a=AutoCluster.autocluster(data,labels=None,clusters=None)

This function automatically applies 13 clustering algorithms, namely, kmeans, kmeans-elkan algorithm, bisectingkmeans,

bisectingkmeans-elkan algorithm, minibatchkmeans, agglomerative, agglomerative-single linkage, agglomerative-complete linkage

agglomerative-average linkage,birch-0.05 threshold,birch-0.1 threshold, birch-0.5 threshold and spectral respectively.

If the labels are not given, only three metrics, namely, the silhouette, davies bouldin and calinski harabasz are used for evaluation.

If the labels are given, the following metrics, namely, 'accuracy','precision','recall','f1','rand','adjusted rand','mutual info',

'adjusted mutual info','fowlkes mallows','homogeneity measure','v measure','silhouette','davies-bouldin','calinski-harabasz','contingency

matrix','confusion matrix' are used for cluster evaluation.

The function finally returns a table that compares the clustering algorithms with the above mentioned metrices.

plot_confusion_matrix

plot_confusion_matrix(a)

This function takes the comparative table as input and plots the confusion matrix for all clustering algorithms.

plot_scatter_plot

plot_scatter_plot(data,a)

This function takes the comparative table as input and plots the scatter plot for all clustering algorithms.

get_metric_plot

get_metric_plot(a,metric)

This function takes the comparative table as input and plot a bar plot of all clustering algorithms based on the given metric.

plot_contingency_matrix

plot_contingency_matrix(a)

This function takes the comparative table as input and plots the contingency matrix for all clustering algorithms.