def cluster(nclus, dfs, n_init=1, variables=None, algorithm='kmeans', ret_gmm=False):
"""
Wrapper around Gaussian Mixture and Kmeans
Parameters:
nclus (int): number of clusters
dfs (pd.DataFrame): the dataframe
n_init (int): the number of times to initialize the clusterer
variables (list): the list of variables to take from the dataframe
algorithm (str): clustering type, `kmeans` or `gmm`
ret_gmm (bool): either `true` for returning the mixture or false to return just the
clustering
Returns:
clusdef +- gmm depending on ret_gmm
"""
from sklearn.cluster import KMeans, hierarchical
from sklearn.mixture import GaussianMixture
# get the data for clustering:
if variables is not None:
nd_pts = dfs[variables].values
else:
nd_pts = dfs
nclus = int(nclus)
if algorithm.lower() == 'kmeans':
clusdef = KMeans(n_clusters=nclus, n_init=n_init).fit(nd_pts).labels_
elif algorithm.lower() == 'gmm':
gmm = GaussianMixture(n_components=nclus, n_init=n_init)
gmm = gmm.fit(nd_pts)
clusdef = np.array(gmm.predict(nd_pts))
elif algorithm.lower() == 'hier':
hier = hierarchical.AgglomerativeClustering(n_clusters=nclus, linkage='ward')
clusdef = hier.fit_predict(nd_pts)
if ret_gmm:
return clusdef, gmm
else:
return clusdef
def a_algo(v, n):
return hierarchical.AgglomerativeClustering(n_clusters=n).fit_predict(v)
h = hclust.linkage(x, method=metoda)
print("Matrice ierarhie:", h, sep="\n")
#Plot ierarhie - graficul dendograma
grafice.dendrograma(h, nume_instante,
"Plot ierarhie. Metoda: " + metoda + " Metrica euclidiana")
#grafice.show()
#Determinare partitie optimala
m = np.shape(h)[
0] #liniile de la shape; shape intoarce nr de linii si nr de coloane
# nr de clustere din partitia optimala - trb sa vad unde este ceas mai mare distanta --ma intereseaza pozitia nu valoarea in sine
k_opt = m - np.argmax(h[:1, 2] - h[:(m - 1), 2])
print("Partitia optimala are " + str(k_opt) + " clusteri")
#di si Pi
# di+1 - d2
#di+1 - di
#cobstruire model sklearn
model_clusterizare_sk = skhclust.AgglomerativeClustering(n_clusters=5,
linkage=metoda)
model_clusterizare_sk.fit(x)
coduri = model_clusterizare_sk.labels_
partitie = np.array(["Cluster" + str(cod + 1) for cod in coduri])
tabel_partitie = pd.DataFrame(data={"Partitie ": partitie},
index=mortalitate.index)
print("Partitia optimala: ", tabel_partitie, sep="\n")
partitie = np.array("Cluster " + str(cod + 1) for cod in coduri)
#Plot partitie in axele principale
grafice.plot_partitie(x, partitie, nume_instante, "Partitie optimala")
grafice.show()
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import scipy.cluster.hierarchy as sch
from sklearn.cluster import hierarchical
from sklearn import datasets
"Import data"
data = pd.DataFrame(datasets.load_iris().data)
shuf = np.random.choice(len(data), size=len(data), replace=False)
y = pd.DataFrame(datasets.load_iris().target)
newData = data.iloc[shuf]
newY = y.iloc[shuf]
"Create dendogram using scipy"
dendogram = sch.dendrogram(sch.linkage(newData, method="ward"))
"Create AgglomerativeClustering model using sklearn"
model = hierarchical.AgglomerativeClustering(n_clusters=3,
linkage="ward",
affinity="euclidean")
"Fit data using created model"
model.fit(newData)
y_pred = model.fit_predict(newData)
"accuracy"
accuracy = sum(list(newY) == y_pred) / len(newY)
print(accuracy)
plt.show()
data = pd.read_csv(r"C:\Users\Bharat Gupta\Downloads\Mall_Customers.csv")
x = data.iloc[0:, [3, 4]]
dendo = sch.dendrogram(
sch.linkage(x, method='ward')
) #ward minimizes the Varience within cluster and KMeans++ minimize WCSS
plt.title('Dendrogram')
plt.xlabel('CUstomers')
plt.ylabel("ED")
plt.show()
#############################################
from sklearn.cluster import hierarchical
model = hierarchical.AgglomerativeClustering(n_clusters=4)
y_sch = model.fit_predict(x)
###########################################
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import scipy.cluster.hierarchy as sch
import seaborn as sns
data = pd.read_csv(r"C:\Users\Bharat Gupta\Desktop\Edge Download\Wine.csv")
x.info()
# data['Malic_Acid']=np.where(data["Malic_Acid"]>5,5,data["Malic_Acid"])
# sns.boxplot(data['Malic_Acid'])
if estimator1.labels_[i]==0:
plt.plot(iris.values[i:,1],iris.values[i:,2],'go')
plt.plot(estimator1.cluster_centers_[:,0],estimator1.cluster_centers_[:,1],'o',c='black')
elif estimator1.labels_[i]==1:
plt.plot(iris.values[i:,1],iris.values[i:,2],'ro')
plt.plot(estimator1.cluster_centers_[:,0],estimator1.cluster_centers_[:,1],'o',c='black')
elif estimator1.labels_[i]==2:
plt.plot(iris.values[i:,1],iris.values[i:,2],'bo')
plt.plot(estimator1.cluster_centers_[:,0],estimator1.cluster_centers_[:,1],'o',c='black')
plt.show()
#Black points are centroids
# # Clustering using Hierarchical Clustering Algorithm
estimator2 = hierarchical.AgglomerativeClustering(n_clusters=3)
estimator2.fit(iris.values[:,1:3])
for i in range(150):
if estimator2.labels_[i]==0:
plt.plot(iris.values[i:,1],iris.values[i:,2],'go')
elif estimator2.labels_[i]==1:
plt.plot(iris.values[i:,1],iris.values[i:,2],'ro')
elif estimator2.labels_[i]==2:
plt.plot(iris.values[i:,1],iris.values[i:,2],'bo')
plt.show()
# # Clustering using DBSCAN Clustering Algorithm
estimator3 = DBSCAN()
def h_agglomerative(values, n):
return hierarchical.AgglomerativeClustering(
n_clusters=n).fit_predict(values)