def inertie_intra(k,X,e,N,perc):
distance=0
for i in range(1,N+1):
Y=labeling(k,X,perc)
centers,clusters=ss_kmeans_pp(Y,k,e)
distance+=SSE(k,Y,centers)
return distance/N
def elbow_manual(n_clusters,X):
sample,features=X.shape
e=10**(-10)
X = DataFrameImputer().fit_transform(X)
SSE=[]
SSE1=[]
for i in range(1,n_clusters):
Y=labeling(i,X,0.6)
centers,labels=ss_kmeans_pp(Y,i,e)
centers_sk,labels_sk=kmean_sklearn(i,X)
# en utilisant lloyd
SSE.append(np.sum(np.min(cdist(X,centers,'euclidean'),axis=1)))
# en utilisant sklearn
SSE1.append(np.sum(np.min(cdist(X,centers_sk,'euclidean'),axis=1)))
K=np.arange(1,n_clusters)
plt.plot(K,SSE,label='méthode manuel',color='blue')
plt.plot(K,SSE1,label='méthode sklearn',color='orange')
plt.xticks(np.arange(1, n_clusters, 1))
kn = KneeLocator(K, SSE1, curve='convex', direction='decreasing')
# plotting dashed_vline on knee
plt.vlines(kn.knee, plt.ylim()[0], plt.ylim()[1], linestyles='dashed')
plt.show()
plt.legend()
# Uploading ML dataset
base = pd.read_csv('BigML_Dataset.csv', sep=',')
X = base.iloc[:, 1:]
# Imputing based on mean for numeric, and most frequent for strings
X = DataFrameImputer().fit_transform(X)
X.fillna(X.mean())
k, e = 5, 10**(-10)
# On choisi le pourcentage unlabeled
fraction = 0.4
#labelisation des données
Y = labeling(k, X, fraction)
#kmeans++ semi supervisés enfin
t0 = time()
centers, clusters = ss_kmeans_pp(Y, k, e)
t1 = time()
print('En utilisant kmeans ++ semi supevisée : %f' %
(t1 - t0)) #max : 0.185409, min : 0.117940, 4 essais
#comparaison avec kmeans++ sklearn
t2 = time()
resultat = kmean_sklearn(k, X)
t3 = time()
print('En utilisant kmeans ++ de sklearn : %f' %
(t3 - t2)) #max : 5.153863, min : 4.261078, 4 essais
X=base.iloc[:,1:]
# Imputing based on mean for numeric, and most frequent for strings
X = DataFrameImputer().fit_transform(X)
X.fillna(X.mean())
quanti_X=X.drop(['shops_used'], axis='columns')
k,e=5,10**(-10)
# On choisi le pourcentage unlabeled
fraction=0.60
#labelisation des données
Y=labeling(k,quanti_X,fraction)
#kmeans++ semi supervisés
t0=time()
centers,clusters = ss_kmeans_pp(Y,k,e)
t1=time()
print('En utilisant kmeans ++ semi supevisée : %f' %(t1-t0)) #max : 0.185409, min : 0.117940, 4 essais
""" Graphique comparaison des clusters"""
# Afficher la variable qualitative: shops_used
# sns.lmplot('min_distance_to_shops', 'products_purchased', data=X, hue='shops_used',fit_reg=False,scatter_kws={"s": 10})
# plt.show()
# plt.title('Distribution de la variable qualitative: shops_used')