数据分析与挖掘（三）：掌握使用库函数进行数据分析的方法

数据分析与挖掘（三）：掌握使用库函数进行数据分析的方法

一、实验目的及要求

掌握使用库函数进行数据分析的方法

掌握K-means算法的代码实现

对比不同聚类算法性能

二、实验设备（环境）及要求

Jupyter notebook、百度 AI studio

三、实验内容

1.使用库函数完成鸢尾花K-means聚类分析。

(1) 实现基本步骤

调用函数代码段：

np.random.seed(5)

iris = datasets.load_iris()

X = iris.data

y = iris.target

est = KMeans(n_clusters=3)

est.fit(X)

labels = est.labels_

显示结果代码段：

fig = plt.figure(1, figsize=(4, 3))

ax = Axes3D(fig, rect=[0, 0, .95, 1], elev=48, azim=134)

ax.scatter(X[:, 3], X[:, 0], X[:, 2],c=labels.astype(np.float), edgecolor='k')

ax.w_xaxis.set_ticklabels([])

ax.w_yaxis.set_ticklabels([])

ax.w_zaxis.set_ticklabels([])

ax.set_xlabel('花瓣宽度')

ax.set_ylabel('萼片长度')

ax.set_zlabel('花瓣长度')

ax.set_title("3类")

ax.dist = 12

plt.show()

(2) 运行（K-means聚类图）

2.使用库函数完成鸢尾花DBSCAN聚类分析。

(1) 实现基本步骤

调用函数代码段：

centers = [[1, 1], [-1, -1], [1, -1]]

X, labels_true = make_blobs(n_samples=750, centers=centers, cluster_std=0.4, random_state=0)

X = StandardScaler().fit_transform(X)

\# Compute DBSCAN

db = DBSCAN(eps=0.3, min_samples=10).fit(X)

core_samples_mask = np.zeros_like(db.labels_, dtype=bool)

core_samples_mask[db.core_sample_indices_] = True

labels = db.labels_

n_clusters_ = len(set(labels)) - (1 if -1 in labels else 0)

 

print('Estimated number of clusters: %d' % n_clusters_)

print("Homogeneity: %0.3f" % metrics.homogeneity_score(labels_true, labels))

print("Completeness: %0.3f" % metrics.completeness_score(labels_true, labels))

print("V-measure: %0.3f" % metrics.v_measure_score(labels_true, labels))

print("Adjusted Rand Index: %0.3f"

   % metrics.adjusted_rand_score(labels_true, labels))

print("Adjusted Mutual Information: %0.3f"

   % metrics.adjusted_mutual_info_score(labels_true, labels))

print("Silhouette Coefficient: %0.3f"

   % metrics.silhouette_score(X, labels))

显示结果代码段：

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 = X[class_member_mask & core_samples_mask]

  plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col),

​       markeredgecolor='k', markersize=14)

 

  xy = X[class_member_mask & ~core_samples_mask]

  plt.plot(xy[:, 0], xy[:, 1], 'o', markerfacecolor=tuple(col),

​       markeredgecolor='k', markersize=6)

 

plt.title('估计类的数量: %d' % n_clusters_)

plt.show()

(2) 运行（DBSCAN聚类图）

3.实现K-means算法，完成鸢尾花聚类分析。

(1) 实现基本步骤

调用函数代码段：

def prepare_data():

  iris = datasets.load_iris()

  x = iris.data

  return x


def k_means(x, k=3):

  \#随机初始化质心

  index_list = np.arange(len(x))

  np.random.shuffle(index_list)

  centroids_index = index_list[:k]

  centroids = x[centroids_index]

  \#初始化标签数组

  y = np.arange(len(x))

  iter_num = 10 #自行设置迭代次数

  \#每次迭代计算新的将现有点分类，计算新的簇中心

  for i in range(iter_num):

​    y_new = np.arange(len(x))

​    for i, xi in enumerate(x):

​      y_new[i] = np.argmin([np.linalg.norm(xi - cj) for cj in centroids])

​    if sum(y != y_new) == 0:

​      break

​    for j in range(k):

​      centroids[j] = np.mean(x[np.where(y_new == j)], axis=0)

​    y = y_new.copy()

  return y

显示结果代码段：

if __name__ == '__main__':

  x=prepare_data()

  k_means(x, k=3)

  \#可视化

  for i in range(np.shape(y)[0]):

​    if y[i] == 0:

​      plt.scatter(x[i][0], x[i][1], c='b', s=20)

​    elif y[i] == 1:

​      plt.scatter(x[i][0], x[i][1], c='y', s=20)

​    else:

​      plt.scatter(x[i][0], x[i][1], c='g', s=20)

  plt.show()

(2) 运行（K-means算法鸢尾花聚类分析图）

4.调用库函数完成鸢尾花聚类分析，对比K-means算法和层次聚类算法性能。

(1) 实现基本步骤

调用函数代码段：

def eva_kmeans(x,y):

  \#运行十次，取平均分类正确样本数，运行时间，准确率

  kmean_ei = 0.0 #分类正确样本数

  kmean_rt = 0.0 #运行时间

  kmean_aa = 0.0 #准确率

  for i in range(0, iter):

​    x_train, x_test, y_train, y_test = train_test_split(x, y, train_size=0.8, test_size=0.2, random_state=1)

​    

​    k_begin = time.time()

​    kmeans = KMeans(init="random", n_clusters=3, random_state=0).fit(x,y)    

​    kmean_pred = kmeans.predict(x_test)

​    k_end = time.time() - k_begin

​    kmean_rt = kmean_rt + k_end

​    

​    accuracy_number = 0

​    for i in range(len(y_test)):

​      if kmean_pred[i] == y_test[i]:

​        accuracy_number += 1

​    kmean_ei = kmean_ei + accuracy_number

​    

​    accuracy_percentage = metrics.accuracy_score(y_test, kmean_pred)*100   

​    kmean_aa = kmean_aa + accuracy_percentage

 

  kmean_ei = kmean_ei / (iter*1.0)

  kmean_rt = kmean_rt / (iter*1.0)

  kmean_aa = kmean_aa / (iter*1.0)

  return kmean_ei, kmean_rt, kmean_aa

 

\#层次聚类

def eva_hierarchical(x,y):

  hier_ei = 0.0

  hier_rt = 0.0

  hier_aa = 0.0

  for i in range(0, 10):

​    x_train, x_test, y_train, y_test = train_test_split(x, y, train_size=0.8, test_size=0.2, random_state=1)

​    

​    k_begin = time.time()

​    hier = AgglomerativeClustering(n_clusters=3, affinity="euclidean", linkage="average").fit(x,y) 

​    hier_pred = hier.fit_predict(x_test)

​    k_end = time.time() - k_begin

​    hier_rt = hier_rt + k_end

​    

​    accuracy_number = 0

​    for i in range(len(y_test)):

​      if hier_pred[i] == y_test[i]:

​        accuracy_number += 1

​    hier_ei = hier_ei + accuracy_number

​    

​    accuracy_percentage = metrics.accuracy_score(y_test, hier_pred)*100

​    hier_aa = hier_aa + accuracy_percentage

  hier_ei = hier_ei / (iter*1.0)

  hier_rt = hier_rt / (iter*1.0)

  hier_aa = hier_aa / (iter*1.0)

  return hier_ei, hier_rt, hier_aa

显示结果代码段：

if __name__ == '__main__':

  x,y=prepare_data()

  kmean_ei, kmean_rt, kmean_aa=eva_kmeans(x,y)

  hier_ei, hier_rt, hier_aa = eva_hierarchical(x, y)

  print ("total iterate:",iter)

  print ("method  ", "平均错误分类样本数量", "平均运算时间/s ", "平均错误比例/%")

  print ("K-means",    kmean_ei,"  ", kmean_rt,"  ", kmean_aa)

​    print ("hierarchical   ", hier_ei, "   ", hier_rt,  "   ", hier_aa)

(2) 运行（图 K-means和层次聚类算法对比）

四、实验结果分析以及出现问题

本次上机我们学习掌握了k均值算法和层次聚类基本原理， 并通过具体实验的对比来理解两种方法之间的联系和区别，利用 K-means 和层次聚类算法分别实现了 UCI 数据集下鸢尾花的聚类分析。通过所选取的度量参数，K-means 算法聚类准确率略高，层次聚类准确率较低，但速度更快。在对精度要求较高的聚类中，使用k 均值算法将会是更优的选择，而对结果需求急切时，则可以使用层次聚类算法，速度将比k均值算法快很多。通过这次上机，对聚类算法有了很多新的了解，通过实战来熟悉聚类算法，这将对以后的学习生活很有帮助。