view Marea/marea_cluster.py @ 21:046ed52017ce draft

Fix on clustering output
author bimib
date Wed, 02 Oct 2019 08:23:19 -0400
parents c71ac0bb12de
children a8825e66c3a0
line wrap: on
line source

# -*- coding: utf-8 -*-
"""
Created on Mon Jun 3 19:51:00 2019

@author: Narger
"""

import sys
import argparse
import os
from sklearn.datasets import make_blobs
from sklearn.cluster import KMeans, DBSCAN, AgglomerativeClustering
from sklearn.metrics import silhouette_samples, silhouette_score, davies_bouldin_score, cluster
import matplotlib.pyplot as plt
import scipy.cluster.hierarchy as shc   
import matplotlib.cm as cm
import numpy as np
import pandas as pd

################################# process args ###############################

def process_args(args):
    parser = argparse.ArgumentParser(usage = '%(prog)s [options]',
                                     description = 'process some value\'s' +
                                     ' genes to create class.')

    parser.add_argument('-ol', '--out_log', 
                        help = "Output log")
    
    parser.add_argument('-in', '--input',
                        type = str,
                        help = 'input dataset')
    
    parser.add_argument('-cy', '--cluster_type',
                        type = str,
                        choices = ['kmeans', 'meanshift', 'dbscan', 'hierarchy'],
                        default = 'kmeans',
                        help = 'choose clustering algorythm')
    
    parser.add_argument('-k1', '--k_min', 
                        type = int,
                        default = 2,
                        help = 'choose minimun cluster number to be generated')
    
    parser.add_argument('-k2', '--k_max', 
                        type = int,
                        default = 7,
                        help = 'choose maximum cluster number to be generated')
    
    parser.add_argument('-el', '--elbow', 
                        type = str,
                        default = 'false',
                        choices = ['true', 'false'],
                        help = 'choose if you want to generate an elbow plot for kmeans')
    
    parser.add_argument('-si', '--silhouette', 
                        type = str,
                        default = 'false',
                        choices = ['true', 'false'],
                        help = 'choose if you want silhouette plots')
    
    parser.add_argument('-db', '--davies', 
                        type = str,
                        default = 'false',
                        choices = ['true', 'false'],
                        help = 'choose if you want davies bouldin scores')
    
    parser.add_argument('-td', '--tool_dir',
                        type = str,
                        required = True,
                        help = 'your tool directory')
                        
    parser.add_argument('-ms', '--min_samples',
                        type = int,
                        help = 'min samples for dbscan (optional)')
                        
    parser.add_argument('-ep', '--eps',
                        type = int,
                        help = 'eps for dbscan (optional)')
    
    
    args = parser.parse_args()
    return args

########################### warning ###########################################

def warning(s):
    args = process_args(sys.argv)
    with open(args.out_log, 'a') as log:
        log.write(s + "\n\n")
    print(s)

########################## read dataset ######################################
    
def read_dataset(dataset):
    try:
        dataset = pd.read_csv(dataset, sep = '\t', header = 0)
    except pd.errors.EmptyDataError:
        sys.exit('Execution aborted: wrong format of dataset\n')
    if len(dataset.columns) < 2:
        sys.exit('Execution aborted: wrong format of dataset\n')
    return dataset

############################ rewrite_input ###################################
    
def rewrite_input(dataset):
    #Riscrivo il dataset come dizionario di liste, 
    #non come dizionario di dizionari
    
    for key, val in dataset.items():
        l = []
        for i in val:
            if i == 'None':
                l.append(None)
            else:
                l.append(float(i))
   
        dataset[key] = l
    
    return dataset

############################## write to csv ##################################
    
def write_to_csv (dataset, labels, name):
    list_labels = labels
    list_values = dataset

    list_values = list_values.tolist()
    d = {'Label' : list_labels, 'Value' : list_values}
    
    df = pd.DataFrame(d, columns=['Value','Label'])

    dest = name + '.tsv'
    df.to_csv(dest, sep = '\t', index = False,
                      header = ['Value', 'Label'])
    
########################### trova il massimo in lista ########################
def max_index (lista):
    best = -1
    best_index = 0
    for i in range(len(lista)):
        if lista[i] > best:
            best = lista [i]
            best_index = i
            
    return best_index
    
################################ kmeans #####################################
    
def kmeans (k_min, k_max, dataset, elbow, silhouette, davies):
    if not os.path.exists('clustering/kmeans_output'):
        os.makedirs('clustering/kmeans_output')
    
        
    if elbow == 'true':
        elbow = True
    else:
        elbow = False
        
    if silhouette == 'true':
        silhouette = True
    else:
        silhouette = False
        
    if davies == 'true':
        davies = True
    else:
        davies = False
        

    range_n_clusters = [i for i in range(k_min, k_max+1)]
    distortions = []
    scores = []
    all_labels = []
    
    for n_clusters in range_n_clusters:
        clusterer = KMeans(n_clusters=n_clusters, random_state=10)
        cluster_labels = clusterer.fit_predict(dataset)
        
        all_labels.append(cluster_labels)
        silhouette_avg = silhouette_score(dataset, cluster_labels)
        scores.append(silhouette_avg)
        distortions.append(clusterer.fit(dataset).inertia_)
        
    best = max_index(scores) + k_min
        
    for i in range(len(all_labels)):
        prefix = ''
        if (i + k_min == best):
            prefix = '_BEST'
            
        write_to_csv(dataset, all_labels[i], 'clustering/kmeans_output/kmeans_with_' + str(i + k_min) + prefix + '_clusters.tsv')
            
        if davies:
            with np.errstate(divide='ignore', invalid='ignore'):
                davies_bouldin = davies_bouldin_score(dataset, all_labels[i])
            warning("\nFor n_clusters = " + str(i + k_min) +
                  " The average davies bouldin score is: " + str(davies_bouldin))
        
       
        if silhouette:
            silihouette_draw(dataset, all_labels[i], i + k_min, 'clustering/kmeans_output/silhouette_with_' + str(i + k_min) + prefix + '_clusters.png')
        
        
    if elbow:
        elbow_plot(distortions, k_min,k_max) 

   
    
    

############################## elbow_plot ####################################
    
def elbow_plot (distortions, k_min, k_max):
    plt.figure(0)
    plt.plot(range(k_min, k_max+1), distortions, marker = 'o')
    plt.xlabel('Number of cluster')
    plt.ylabel('Distortion')
    s = 'clustering/kmeans_output/elbow_plot.png'
    fig = plt.gcf()
    fig.set_size_inches(18.5, 10.5, forward = True)
    fig.savefig(s, dpi=100)
    
    
############################## silhouette plot ###############################
def silihouette_draw(dataset, labels, n_clusters, path):
    silhouette_avg = silhouette_score(dataset, labels)
    warning("For n_clusters = " + str(n_clusters) +
          " The average silhouette_score is: " + str(silhouette_avg))
           
    plt.close('all')
    # Create a subplot with 1 row and 2 columns
    fig, (ax1) = plt.subplots(1, 1)
    
    fig.set_size_inches(18, 7)
        
    # The 1st subplot is the silhouette plot
    # The silhouette coefficient can range from -1, 1 but in this example all
    # lie within [-0.1, 1]
    ax1.set_xlim([-1, 1])
    # The (n_clusters+1)*10 is for inserting blank space between silhouette
    # plots of individual clusters, to demarcate them clearly.
    ax1.set_ylim([0, len(dataset) + (n_clusters + 1) * 10])
    
    # Compute the silhouette scores for each sample
    sample_silhouette_values = silhouette_samples(dataset, labels)
        
    y_lower = 10
    for i in range(n_clusters):
        # Aggregate the silhouette scores for samples belonging to
        # cluster i, and sort them
        ith_cluster_silhouette_values = \
        sample_silhouette_values[labels == i]
        
        ith_cluster_silhouette_values.sort()
    
        size_cluster_i = ith_cluster_silhouette_values.shape[0]
        y_upper = y_lower + size_cluster_i
    
        color = cm.nipy_spectral(float(i) / n_clusters)
        ax1.fill_betweenx(np.arange(y_lower, y_upper),
                          0, ith_cluster_silhouette_values,
                                     facecolor=color, edgecolor=color, alpha=0.7)
        
        # Label the silhouette plots with their cluster numbers at the middle
        ax1.text(-0.05, y_lower + 0.5 * size_cluster_i, str(i))
        
        # Compute the new y_lower for next plot
        y_lower = y_upper + 10  # 10 for the 0 samples
    
        ax1.set_title("The silhouette plot for the various clusters.")
        ax1.set_xlabel("The silhouette coefficient values")
        ax1.set_ylabel("Cluster label")
        
        # The vertical line for average silhouette score of all the values
        ax1.axvline(x=silhouette_avg, color="red", linestyle="--")
    
        ax1.set_yticks([])  # Clear the yaxis labels / ticks
        ax1.set_xticks([-0.1, 0, 0.2, 0.4, 0.6, 0.8, 1])
        
        
        plt.suptitle(("Silhouette analysis for clustering on sample data "
                          "with n_clusters = " + str(n_clusters) + "\nAverage silhouette_score = " + str(silhouette_avg)), fontsize=12, fontweight='bold')
            
            
        plt.savefig(path, bbox_inches='tight')
            
######################## dbscan ##############################################
    
def dbscan(dataset, eps, min_samples):
    if not os.path.exists('clustering/dbscan_output'):
        os.makedirs('clustering/dbscan_output')
        
    if eps is not None:
    	clusterer = DBSCAN(eps = eps, min_samples = min_samples)
    else:
    	clusterer = DBSCAN()
    
    clustering = clusterer.fit(dataset)
    
    core_samples_mask = np.zeros_like(clustering.labels_, dtype=bool)
    core_samples_mask[clustering.core_sample_indices_] = True
    labels = clustering.labels_

    # Number of clusters in labels, ignoring noise if present.
    n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)
    
    silhouette_avg = silhouette_score(dataset, labels)
    warning("For n_clusters =" + str(n_clusters_) + 
              "The average silhouette_score is :" + str(silhouette_avg))
    
    ##TODO: PLOT SU DBSCAN (no centers) e HIERARCHICAL

    # Black removed and is used for noise instead.
    unique_labels = set(labels)
    colors = [plt.cm.Spectral(each)
          for each in np.linspace(0, 1, len(unique_labels))]
    for k, col in zip(unique_labels, colors):
        if k == -1:
            # Black used for noise.
            col = [0, 0, 0, 1]

        class_member_mask = (labels == k)
    
        xy = dataset[class_member_mask & core_samples_mask]
        plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col),
                 markeredgecolor='k', markersize=14)
    
        xy = dataset[class_member_mask & ~core_samples_mask]
        plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col),
                 markeredgecolor='k', markersize=6)

    plt.title('Estimated number of clusters: %d' % n_clusters_)
    s = 'clustering/dbscan_output/dbscan_plot.png'
    fig = plt.gcf()
    fig.set_size_inches(18.5, 10.5, forward = True)
    fig.savefig(s, dpi=100)
    
    
    write_to_csv(dataset, labels, 'clustering/dbscan_output/dbscan_results.tsv')
    
########################## hierachical #######################################
    
def hierachical_agglomerative(dataset, k_min, k_max):

    if not os.path.exists('clustering/agglomerative_output'):
        os.makedirs('clustering/agglomerative_output')
    
    plt.figure(figsize=(10, 7))  
    plt.title("Customer Dendograms")  
    shc.dendrogram(shc.linkage(dataset, method='ward'))  
    fig = plt.gcf()
    fig.savefig('clustering/agglomerative_output/dendogram.png', dpi=200)
    
    range_n_clusters = [i for i in range(k_min, k_max+1)]

    for n_clusters in range_n_clusters:
        
        cluster = AgglomerativeClustering(n_clusters=n_clusters, affinity='euclidean', linkage='ward')  
        cluster.fit_predict(dataset)  
        cluster_labels = cluster.labels_
        
        silhouette_avg = silhouette_score(dataset, cluster_labels)
        warning("For n_clusters =", n_clusters,
              "The average silhouette_score is :", silhouette_avg)
        
        plt.clf()
        plt.figure(figsize=(10, 7))  
        plt.title("Agglomerative Hierarchical Clustering\nwith " + str(n_clusters) + " clusters and " + str(silhouette_avg) + " silhouette score")
        plt.scatter(dataset[:,0], dataset[:,1], c = cluster_labels, cmap='rainbow') 
        s = 'clustering/agglomerative_output/hierachical_' + str(n_clusters) + '_clusters.png'
        fig = plt.gcf()
        fig.set_size_inches(10, 7, forward = True)
        fig.savefig(s, dpi=200)
        
        write_to_csv(dataset, cluster_labels, 'clustering/agglomerative_output/agglomerative_hierarchical_with_' + str(n_clusters) + '_clusters.tsv')
        
       

    
############################# main ###########################################


def main():
    if not os.path.exists('clustering'):
        os.makedirs('clustering')

    args = process_args(sys.argv)
    
    #Data read
    
    X = read_dataset(args.input)
    X = pd.DataFrame.to_dict(X, orient='list')
    X = rewrite_input(X)
    X = pd.DataFrame.from_dict(X, orient = 'index')
    
    for i in X.columns:
        tmp = X[i][0]
        if tmp == None:
            X = X.drop(columns=[i])
                
    X = pd.DataFrame.to_numpy(X)
    
    
    if args.cluster_type == 'kmeans':
        kmeans(args.k_min, args.k_max, X, args.elbow, args.silhouette, args.davies)
    
    if args.cluster_type == 'dbscan':
        dbscan(X, args.eps, args.min_samples)
        
    if args.cluster_type == 'hierarchy':
        hierachical_agglomerative(X, args.k_min, args.k_max)
        
##############################################################################

if __name__ == "__main__":
    main()