python数据挖掘（从数据集中抽取特征）

import numpy as np

x= np.arange(30).reshape((10,3))#10个个体、3个特征的数据集
print(x)
x[:,1] = 1 #把所有第二列的数值都改为1
print(x)
from sklearn.feature_selection import VarianceThreshold
vt = VarianceThreshold() #VarianceThreshold转换器，用它处理数据集
Xt = vt.fit_transform(x)
print(Xt)#第二列消失
print(vt.variances_)#输出每一列的方差

结果：
[[ 0  1  2]
 [ 3  4  5]
 [ 6  7  8]
 [ 9 10 11]
 [12 13 14]
 [15 16 17]
 [18 19 20]
 [21 22 23]
 [24 25 26]
 [27 28 29]]
[[ 0  1  2]
 [ 3  1  5]
 [ 6  1  8]
 [ 9  1 11]
 [12  1 14]
 [15  1 17]
 [18  1 20]
 [21  1 23]
 [24  1 26]
 [27  1 29]]
[[ 0  2]
 [ 3  5]
 [ 6  8]
 [ 9 11]
 [12 14]
 [15 17]
 [18 20]
 [21 23]
 [24 26]
 [27 29]]
[ 74.25   0.    74.25]

import os
import pandas as pd
data_folder =  os.path.join(os.getcwd(),‘Data‘,‘adult‘)
adult_filename = os.path.join(data_folder,‘adult.data.txt‘)
adult = pd.read_csv(adult_filename,header=None,
                    names=["Age", "Work-Class", "fnlwgt",
                           "Education", "Education-Num",
                           "Marital-Status", "Occupation",
                           "Relationship", "Race", "Sex",
                           "Capital-gain", "Capital-loss",
                           "Hours-per-week", "Native-Country",
                           "Earnings-Raw"])
adult.dropna(how=‘all‘, inplace=True) #我们需要删除包含无效数字的行（设置inplace参数为真，表示改动当前数据框，而不是新建一个）。
# print(adult["Work-Class"].unique())#数据框的unique函数就能得到所有的工作情况
adult["LongHours"] = adult["Hours-per-week"] > 40 #通过离散化过程转换为类别型特征,把连续值转换为类别型特征

#测试单个特征在Adult数据集上的表现,
X = adult[["Age", "Education-Num", "Capital-gain", "Capital-loss","Hours-per-week"]].values
y = (adult["Earnings-Raw"] == ‘ >50K‘).values
from sklearn.feature_selection import SelectKBest
from sklearn.feature_selection import chi2
transformer = SelectKBest(score_func=chi2, k=3) #使用SelectKBest转换器类，用卡方函数打分，初始化转换器
Xt_chi2 = transformer.fit_transform(X, y)#调用fit_transform方法，对相同的数据集进行预处理和转换
print(transformer.scores_)#每一列的相关性
from sklearn.tree import DecisionTreeClassifier
from sklearn.cross_validation import cross_val_score
clf = DecisionTreeClassifier(random_state=14)
scores_chi2 = cross_val_score(clf, Xt_chi2, y, scoring=‘accuracy‘)
print(scores_chi2)
结果：

[ 8.60061182e+03 2.40142178e+03 8.21924671e+07 1.37214589e+06
6.47640900e+03]
[ 0.82577851 0.82992445 0.83009306] #正确率达到83%

from collections import defaultdict
import os
import numpy as np
import pandas as pd
data_folder = os.path.join(os.getcwd(), "Data")
data_filename = os.path.join(data_folder, "adult", "ad.data.txt")
#前几个特征是数值，但是pandas会把它们当成字符串。要修复这个问题，我们需要编写将字符串转换为数字的函数，该函数能够把只包含数字的字符串转换为数字，把其余的转化为“NaN”
def convert_number(x):
    try:
        return float(x)
    except ValueError:
        return np.nan
converters = defaultdict(convert_number)
converters[1558] = lambda x: 1 if x.strip() == "ad." else 0 #把类别这一列各个类别值由字符串转换为数值
for i in range(1558):#要这样定义才使得字典前面有定义
    converters[i]=lambda x:convert_number(x)
ads = pd.read_csv(data_filename, header=None, converters=converters)
# print(ads[:5])
ads.dropna(inplace=True)#删除空行
#抽取用于分类算法的x矩阵和y数组
X = ads.drop(1558, axis=1).values
y = ads[1558]
from sklearn.decomposition import PCA #主成分分析算法（Principal Component Analysis，PCA）的目的是找到能用较少信息描述数据集的特征组合，用PCA算法得到的数据创建模型，不仅能够近似地表示原始数据集，还能提升分类任务的正确率
pca = PCA(n_components=5)
Xd = pca.fit_transform(X)
np.set_printoptions(precision=3, suppress=True)
print(pca.explained_variance_ratio_ )#每个特征的方差
from sklearn.tree import DecisionTreeClassifier
from sklearn.cross_validation import cross_val_score
clf = DecisionTreeClassifier(random_state=14)
scores_reduced = cross_val_score(clf, Xd, y, scoring=‘accuracy‘)
print(scores_reduced)

#把PCA返回的前两个特征做成图形
from matplotlib import pyplot as plt
classes = set(y)
colors = [‘red‘, ‘green‘]
for cur_class, color in zip(classes, colors):
    mask = (y == cur_class).values
    plt.scatter(Xd[mask, 0], Xd[mask, 1], marker=‘o‘, color=color,
                label=int(cur_class))

plt.legend()
plt.show()

结果：
[ 0.854  0.145  0.001  0.     0.   ]
[ 0.944  0.924  0.925]