Pythonではじめる教師なし学習 ―機械学習の可能性を広げるラベルなしデータの利用

5.3

 

k

平均法

125

# k-means - Accuracy as the number of components varies

n_clusters = 20

n_init = 10

max_iter = 300

tol = 0.0001

random_state = 2018

n_jobs = 2

kMeans_inertia = pd.DataFrame(data=[],index=[9, 49, 99, 199, \

299, 399, 499, 599, 699, 783],columns=['inertia'])

overallAccuracy_kMeansDF = pd.DataFrame(data=[],index=[9, 49, \

99, 199, 299, 399, 499, 599, 699, 783], \

columns=['overallAccuracy'])

for cutoffNumber in [9, 49, 99, 199, 299, 399, 499, 599, 699, 783]:

kmeans = KMeans(n_clusters=n_clusters, n_init=n_init, \

max_iter=max_iter, tol=tol, random_state=random_state, \

n_jobs=n_jobs)

cutoff = cutoffNumber

kmeans.fit(X_train_PCA.loc[:,0:cutoff]) ...