此处基于Bisecting K-means算法处理Iris数据集
bisecting_kmeans.py模块：
import numpy as np
class KMeansClassifier():
"""初始化KMeansClassifier类"""
def __init__(self, k=3, initCent='random', max_iter=500):
# 类的成员数据(变量前用下划线)
self._k = k # 中心点
self._initCent = initCent # 生成初始中心点
self._max_iter = max_iter # 最大迭代次数
self._clusterAssment = None # 点分配结果
self._labels = None
self._sse = None # 误差平方和
def _calEDist(self, arrA, arrB):
"""计算欧氏距离，参数为两个一维数组"""
return np.math.sqrt(sum(np.power(arrA-arrB, 2)))
def _calMDist(self, arrA, arrB):
"""计算曼哈顿距离，参数为两个一维数组"""
return sum(np.abs(arrA-arrB))
def _randCent(self, data_X, k):
"""随机选取k个质心，返回一个k*n的质心矩阵"""
n = data_X.shape[1] # 特征的维度
centroids = np.empty((k,n)) # 使用numpy生成一个k*n的矩阵，用于存储质心
for j in range(n):
minJ = min(data_X[:, j])
rangeJ = float(max(data_X[:, j] - minJ))
centroids[:, j] = (minJ + rangeJ * np.random.rand(k, 1)).flatten() # 使用flatten函数展平嵌套列表(nested list)
return centroids
def fit(self, data_X):
"""参数为m*n维矩阵"""
if not isinstance(data_X, np.ndarray):
data_X = np.asarray(data_X)
m = data_X.shape[0] # 样本的个数
self._clusterAssment = np.zeros((m,2)) # 一个m*2维矩阵，矩阵第一列存储样本点所属的簇的索引值，第二列存储该点与所属簇的质心的平方误差
if self._initCent == 'random':
self._centroids = self._randCent(data_X, self._k)
clusterChanged = True
for _ in range(self._max_iter):
clusterChanged = False
for i in range(m): # 将每个样本点分配到离它最近的质心所属的簇
minDist = np.inf # 首先将minDist置为一个无穷大的数
minIndex = -1 # 将最近质心的下标置为-1
for j in range(self._k): # k次迭代用于寻找最近的质心
arrA = self._centroids[j,:]
arrB = data_X[i,:]
distJI = self._calEDist(arrA, arrB) # 计算误差值
if distJI < minDist:
minDist = distJI
minIndex = j
if self._clusterAssment[i, 0] != minIndex or self._clusterAssment[i, 1] > minDist**2:
clusterChanged = True
self._clusterAssment[i,:] = minIndex, minDist**2
if not clusterChanged: # 若所有样本点所属的簇都不改变,则已收敛,结束迭代
break
# 更新质心，将每个簇中的点的均值作为质心
for i in range(self._k):
index_all = self._clusterAssment[:,0] # 取出样本所属簇的索引值
value = np.nonzero(index_all==i) # 取出所有属于第i个簇的索引值
ptsInClust = data_X[value[0]] # 取出属于第i个簇的所有样本点
self._centroids[i,:] = np.mean(ptsInClust, axis=0) # 计算均值
self._labels = self._clusterAssment[:,0]
self._sse = sum(self._clusterAssment[:,1])
def predict(self, X):
"""根据聚类结果，预测新输入数据所属的族"""
if not isinstance(X,np.ndarray):
X = np.asarray(X)
m = X.shape[0] # m代表样本数量
preds = np.empty((m,))
# 将每个样本点分配到离它最近的质心所属的族
for i in range(m):
minDist = np.inf
for j in range(self._k):
distJI = self._calEDist(self._centroids[j,:], X[i,:])
if distJI < minDist:
minDist = distJI
preds[i] = j
return preds
class biKMeansClassifier():
"""初始化biKMeansClassifier类"""
def __init__(self, k=3):
self._k = k
self._centroids = None
self._clusterAssment = None
self._labels = None
self._sse = None
def _calEDist(self, arrA, arrB):
"""计算欧氏距离，参数为两个一维数组"""
return np.math.sqrt(sum(np.power(arrA-arrB, 2)))
def fit(self, X):
m = X.shape[0]
self._clusterAssment = np.zeros((m,2))
# 创建初始簇
centroid0 = np.mean(X, axis=0).tolist()
centList =[centroid0]
for j in range(m): # 计算每个样本点与质心之间初始的平方误差
self._clusterAssment[j,1] = self._calEDist(np.asarray(centroid0), X[j,:])**2
while (len(centList) < self._k):
lowestSSE = np.inf
# 尝试划分每一簇,选取使得误差最小的那个簇进行划分
for i in range(len(centList)):
index_all = self._clusterAssment[:,0] # 取出样本所属簇的索引值
value = np.nonzero(index_all==i) # 取出所有属于第i个簇的索引值
ptsInCurrCluster = X[value[0],:] # 取出属于第i个簇的所有样本点
clf = KMeansClassifier(k=2)
clf.fit(ptsInCurrCluster)
# 划分该簇后，所得到的质心、分配结果及误差矩阵
centroidMat, splitClustAss = clf._centroids, clf._clusterAssment
sseSplit = sum(splitClustAss[:,1])
index_all = self._clusterAssment[:,0]
value = np.nonzero(index_all==i)
sseNotSplit = sum(self._clusterAssment[value[0],1])
if (sseSplit + sseNotSplit) < lowestSSE:
bestCentToSplit = i
bestNewCents = centroidMat
bestClustAss = splitClustAss.copy()
lowestSSE = sseSplit + sseNotSplit
# 该簇被划分成两个子簇后,其中一个子簇的索引变为原簇的索引
# 另一个子簇的索引变为len(centList),然后存入centList
bestClustAss[np.nonzero(bestClustAss[:,0]==1)[0],0]=len(centList)
bestClustAss[np.nonzero(bestClustAss[:,0]==0)[0],0]=bestCentToSplit
centList[bestCentToSplit] = bestNewCents[0,:].tolist()
centList.append(bestNewCents[1,:].tolist())
self._clusterAssment[np.nonzero(self._clusterAssment[:,0] == bestCentToSplit)[0],:]= bestClustAss
self._labels = self._clusterAssment[:,0]
self._sse = sum(self._clusterAssment[:,1])
self._centroids = np.asarray(centList)
def predict(self, X):
"""根据聚类结果，预测新输入数据所属的族"""
if not isinstance(X,np.ndarray):
X = np.asarray(X)
m = X.shape[0] # m代表样本数量
preds = np.empty((m,))
for i in range(m): # 将每个样本点分配到离它最近的质心所属的族
minDist = np.inf
for j in range(self._k):
distJI = self._calEDist(self._centroids[j,:],X[i,:])
if distJI < minDist:
minDist = distJI
preds[i] = j
return preds
test.py模块：
import pandas as pd
import numpy as np
from bisecting_kmeans import biKMeansClassifier
import matplotlib.pyplot as plt
if __name__=="__main__":
data_X = pd.read_csv(r"iris.csv")
data_X = data_X.drop(data_X.columns[4], axis=1)
data_X = np.array(data_X)
# print(data_X)
# k = 2
k = 3
clf = biKMeansClassifier(k) # 实例化biKMeansClassifier类
clf.fit(data_X) # 训练模型
cents = clf._centroids
labels = clf._labels
sse = clf._sse
colors = ['b','r','m']
for i in range(k):
index = np.nonzero(labels==i)[0]
# print(index)
x0 = data_X[index, 0]
x1 = data_X[index, 1]
y_i = i
for j in range(len(x0)):
plt.scatter(x0[j], x1[j], color=colors[i])
plt.scatter(cents[i,0], cents[i,1], marker='x', color=colors[i], linewidths=7)
# plt.title("SSE={:.2f}".format(sse))
print("误差平方和为: ", sse)
plt.axis([4.2,8.2,1.8,4.6])
outname = "./bisecting_kmeans" + str(k) + ".png"
plt.savefig(outname)
plt.show()
Output：
当K选取为2时：
参考资料：《机器学习实战》
