利用sklearn中的Kmeans对seeds_dataset数据集进行分类分析

---

前言

本篇文章是主要讲述利用sklearn库中的Kmeans方法进行分类，对于k-means算法的原理不做过多介绍

---

一、k-means主要步骤

step1：选定要聚类的类别数目k（如上例的k=3类），选择k个中心点。

step2：针对每个样本点，找到距离其最近的中心点（寻找组织），距离同一中心点最近的点为一个类，这样完成了一次聚类。

step3：判断聚类前后的样本点的类别情况是否相同，如果相同，则算法终止，否则进入step4。

step4：针对每个类别中的样本点，计算这些样本点的中心点，当做该类的新的中心点，继续step2。

上述是实现K-means的基本算法
而sklearn中有Kmeans库就就很方便了

二、数据集

这里有数据集的具体介绍->下载地址
数据和代码也可以到我的GitHub上下载：github

三、不使用PCA降维

1.读入数据

代码如下：

from pandas import read_csv
import numpy as np
from sklearn import metrics
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt

filename='D:\datasets\mldata\seeds_dataset.csv' #地址 
names = ['area','perimeter','compactness','klength','width','asymmetry','kglength','kind']
re = read_csv(filename,names=names)   #读取数据
seed = np.array(re)  #转换为numpy数组
#print(seed.shape)
np.random.shuffle(seed)  #打乱顺序
X_train = seed[:180:,0:6]   #分割数据集
X_test = seed[180:,0:6]

2.找簇心

distortions = []
for i in range(1, 40):
        km = KMeans(n_clusters=i,
                    init='k-means++',
                    n_init=10,
                    max_iter=300,
                    random_state=0)
        km.fit(X_train)
        distortions.append(km.inertia_)

plt.plot(range(1,40), distortions, marker='o')
plt.xlabel('Number of clusters')
plt.ylabel('Distortion')
plt.show()

找簇心可以参考这篇博客：传送门

3.训练以及评估

代码如下：

kmeans = KMeans(n_clusters=2, random_state=10)  #定义三个质心，并拟合质心
kmeans.fit(X_train)

#print(kmeans.score(X_test))
#print(kmeans.labels_)   #标签 
print("簇心={}".format(kmeans.cluster_centers_))  #打印簇心
print("测试结果：{}".format(kmeans.predict(X_test)))  #预测样本
print("CH分数={}".format(metrics.calinski_harabasz_score(X_train, kmeans.labels_))) 
print("轮廓系数={}".format(metrics.silhouette_score(X_train,kmeans.labels_,metric='euclidean'))) 

4.完整代码

from pandas import read_csv
import numpy as np
from sklearn import metrics
from sklearn.cluster import KMeans
import matplotlib.pyplot as plt
filename='D:\datasets\mldata\seeds_dataset.csv' #地址 
names = ['area','perimeter','compactness','klength','width','asymmetry','kglength','kind']
re = read_csv(filename,names=names)   #读取数据
seed = np.array(re)  #转换为numpy数组
#print(seed.shape)
np.random.shuffle(seed)  #打乱顺序
X_train = seed[:180:,0:6]   #分割数据集
X_test = seed[180:,0:6]
#print(X_train.shape)
#print(X_test.shape)

distortions = []
for i in range(1, 40):
        km = KMeans(n_clusters=i,
                    init='k-means++',
                    n_init=10,
                    max_iter=300,
                    random_state=0)
        km.fit(X_train)
        distortions.append(km.inertia_)

plt.plot(range(1,40), distortions, marker='o')
plt.xlabel('Number of clusters')
plt.ylabel('Distortion')
plt.show()

kmeans = KMeans(n_clusters=2, random_state=10)  #定义三个质心，并拟合质心
kmeans.fit(X_train)

#print(kmeans.score(X_test))
#print(kmeans.labels_)   #标签 
print("簇心={}".format(kmeans.cluster_centers_))  #打印簇心
print("测试结果：{}".format(kmeans.predict(X_test)))  #预测样本
print("CH分数={}".format(metrics.calinski_harabasz_score(X_train, kmeans.labels_))) 
print("轮廓系数={}".format(metrics.silhouette_score(X_train,kmeans.labels_,metric='euclidean'))) 

---

四、使用PCA降维

代码如下：

from pandas import read_csv
import numpy as np
from sklearn import metrics
import matplotlib.pyplot as plt
from sklearn.cluster import KMeans
from sklearn.decomposition import PCA

filename='D:\datasets\mldata\seeds_dataset.csv' #地址 
names = ['area','perimeter','compactness','klength','width','asymmetry','kglength','kind']
re = read_csv(filename,names=names)   #读取数据
seed = np.array(re)  #转换为numpy数组
#print(seed.shape)
np.random.shuffle(seed)  #打乱顺序
X_train = seed[:180:,0:6]   #训练数据
X_test = seed[180:,0:6]
#print(X_train.shape)
#print(X_test.shape)

pca = PCA(0.95)
pca.fit(X_train)
X_train_reduction = pca.transform(X_train)
X_test_reduction = pca.transform(X_test)
#print(X_train_reduction.shape)

kmeans = KMeans(n_clusters=2, random_state=10)  #定义三个质心，并拟合质心
kmeans.fit(X_train_reduction)
#print(seed[:,7])
#print(kmeans.score(X_test))
print("簇心={}".format(kmeans.cluster_centers_))  #打印簇心
print("测试结果：{}".format(kmeans.predict(X_test_reduction)))  #预测样本
print("CH分数={}".format(metrics.calinski_harabasz_score(X_train_reduction, kmeans.labels_))) 
print("轮廓系数={}".format(metrics.silhouette_score(X_train_reduction,kmeans.labels_,metric='euclidean'))) 

plt.scatter(X_train_reduction[:,0],X_train_reduction[:,1],c=kmeans.labels_)
plt.show()

五、对比结果

不使用PCA:

使用PCA：

结论：降维后对准确率影响并不大