3-2 特征二值化

特征二值化

请参考《数据准备和特征工程》中的相关章节，调试如下代码。

---

基础知识

import pandas as pd

pm25 = pd.read_csv("/home/aistudio/data/data20505/pm2.csv")
pm25.head()

	RANK	CITY_ID	CITY_NAME	Exposed days
0	1	594	拉萨	2
1	2	579	玉溪	7
2	3	263	厦门	8
3	4	267	泉州	9
4	5	271	漳州	10

import numpy as np

# 以平均值作为阈值对特征"Exposed days"进行二值化
pm25['bdays'] = np.where(pm25["Exposed days"] > pm25["Exposed days"].mean(), 1, 0)
pm25.sample(10)

	RANK	CITY_ID	CITY_NAME	Exposed days	bdays
174	197	598	宝鸡	128	1
260	283	364	安阳	246	1
137	152	183	南通	103	0
249	272	358	开封	210	1
54	62	44	朔州	58	0
217	240	315	东营	170	1
86	96	86	抚顺	78	0
77	86	303	上饶	71	0
190	213	608	商洛	138	1
180	203	438	岳阳	132	1

from sklearn.preprocessing import Binarizer

# 创建二值化模型,并用平均值赋值
bn = Binarizer(threshold=pm25["Exposed days"].mean())   

# 用得到的模型进行训练并同时实现特征转换,pm25[["Exposed days"]将Series转换为DataFrame对象
print(type(pm25["Exposed days"]))
result = bn.fit_transform(pm25[["Exposed days"]])   

pm25['sk-bdays'] = result
pm25.sample(10)

	RANK	CITY_ID	CITY_NAME	Exposed days	bdays	sk-bdays
92	103	147	双鸭山	82	0	0
12	13	273	南平	18	0	0
175	198	545	遂宁	129	1	1
157	179	449	怀化	119	1	1
67	75	507	钦州	65	0	0
40	45	88	丹东	49	0	0
41	46	462	珠海	49	0	0
121	133	230	衢州	94	0	0
70	78	510	玉林	67	0	0
176	199	199	泰州	130	1	1

pm25[["Exposed days"]].shape

(264, 1)

pm25["Exposed days"].shape

(264,)

pm25["Exposed days"].values.reshape((-1, 1)).shape

(264, 1)

from sklearn.preprocessing import binarize

# 训练出的模型fbin可以对其他DataFrame对象进行二值化操作
fbin = binarize(pm25[['Exposed days']], threshold=pm25['Exposed days'].mean())

fbin[[1, 50, 100, 150, 200]]

array([[0],
       [0],
       [0],
       [1],
       [1]])

# scale(float)：正态分布的标准差，越大曲线越矮胖
gau = np.random.normal(loc=0, scale=1.0, size=100)
gau

array([-0.16138569, -1.12381876,  1.08345071, -0.3374515 , -0.4377176 ,
       -0.18485122, -0.47717794, -0.15147513,  1.34975203, -0.06388386,
        1.2794776 , -0.67413457, -2.03388881,  1.77891998,  3.45445178,
       -0.93258988,  0.39723041, -1.23677885,  1.87841988, -0.48846415,
        0.23898558,  0.08322678,  0.50841094, -0.59189042, -0.86218771,
        0.13808454, -1.420791  , -0.52815037,  0.37716549,  0.55944191,
       -0.81171679,  0.26489442, -0.62432789, -0.43654577,  0.226915  ,
       -1.01346821, -1.42727242, -0.45127134, -0.18215018,  1.11537106,
        0.30099939, -1.2661621 , -0.11895918,  1.69860201, -0.67702066,
       -1.00599679, -0.69448062, -0.94056253, -0.25849202, -0.08299086,
        1.48469908, -1.19206442,  0.95998195, -1.800488  , -0.04798554,
       -0.77909029,  0.9659936 , -0.38096705,  1.52306246, -2.033555  ,
       -1.82498521, -0.4745455 , -0.20866822,  2.01935722,  0.46819346,
        0.37152816, -0.20247084, -1.14168624,  1.04413851,  0.98376221,
        0.27129983, -0.66495964, -0.99604697,  0.31477433,  1.14606679,
        0.92117707,  0.91663896,  0.96625631, -2.00554469,  1.02536304,
       -0.63002324,  1.71252177, -0.65706596, -1.33159033,  0.08011075,
        1.62804803, -1.63617324,  1.42729399, -2.14112983,  0.95559999,
       -0.74515346,  1.29242505,  0.03208948, -0.45625835,  1.24445081,
       -1.53939509,  0.40075234, -0.97061926,  0.39624106, -0.14267309])

# Binarizer()的参数默认为0，reshape(-1, 1)中-1表示按照行的方式的获取，结果形成一列
gau_bin = Binarizer().fit_transform(gau.reshape(-1, 1))   
gau_bin.reshape(1,-1) [0]

array([0., 0., 1., 0., 0., 0., 0., 0., 1., 0., 1., 0., 0., 1., 1., 0., 1.,
       0., 1., 0., 1., 1., 1., 0., 0., 1., 0., 0., 1., 1., 0., 1., 0., 0.,
       1., 0., 0., 0., 0., 1., 1., 0., 0., 1., 0., 0., 0., 0., 0., 0., 1.,
       0., 1., 0., 0., 0., 1., 0., 1., 0., 0., 0., 0., 1., 1., 1., 0., 0.,
       1., 1., 1., 0., 0., 1., 1., 1., 1., 1., 0., 1., 0., 1., 0., 0., 1.,
       1., 0., 1., 0., 1., 0., 1., 1., 0., 1., 0., 1., 0., 1., 0.])

项目案例

%matplotlib inline
import matplotlib.pyplot as plt
import cv2

# 写一个专门在Jupyter中显示图片的函数
def show_img(img):    
    if len(img.shape) == 3:

        # 分离得到各个通道的灰度值(单通道图像)
        b, g, r = cv2.split(img)

        # 合并单通道成多通道(不能合并多个多通道图像)。   
        img = cv2.merge([r, g, b])
        plt.imshow(img)
    else:
        plt.imshow(img, cmap="gray")
    plt.axis("off")
    plt.show()

laoqi = cv2.imread("work/images/laoqi.png")
show_img(laoqi)

# BGR转GRAY
gray_laoqi = cv2.cvtColor(laoqi, cv2.COLOR_BGR2GRAY)
show_img(gray_laoqi)

# cv2.threshold(img, threshold, maxval,type)
# threshold是设定的阈值127
# maxval是当灰度值大于（或小于）阈值127时将该灰度值赋成的值255
# type规定的是当前二值化的方式  ，cv2.THRESH_BINARY：大于阈值127的部分被置为255，小于部分被置为0    
ret,thr = cv2.threshold(gray_laoqi, 127, 255, cv2.THRESH_BINARY)
show_img(thr)

动手练习

pd.read_csv("/home/aistudio/data/data20512/marathon.csv").head()

	age	gender	split	final
0	33	M	01:05:38	02:08:51
1	32	M	01:06:26	02:09:28
2	31	M	01:06:49	02:10:42
3	38	M	01:06:16	02:13:45
4	31	M	01:06:32	02:13:59

import datetime

# 将时间换算为datetime.timedelta类型，split代表半马用时
def convert_time(s):
    h,m,s = map(int, s.split(":"))
    return datetime.timedelta(hours=h, minutes=m, seconds=s)

marathon = pd.read_csv("/home/aistudio/data/data20512/marathon.csv",
                      converters={"split":convert_time, 
                                  "final":convert_time})
print(marathon.dtypes)
marathon.head()

age                 int64
gender             object
split     timedelta64[ns]
final     timedelta64[ns]
dtype: object

	age	gender	split	final
0	33	M	0 days 01:05:38	0 days 02:08:51
1	32	M	0 days 01:06:26	0 days 02:09:28
2	31	M	0 days 01:06:49	0 days 02:10:42
3	38	M	0 days 01:06:16	0 days 02:13:45
4	31	M	0 days 01:06:32	0 days 02:13:59

# 由于转换为int后的存储单位是纳秒，因此需要乘以1e-9
marathon['split'] = marathon['split'].astype(int) * 1e-9
marathon['final'] = marathon['final'].astype(int) * 1e-9

marathon.head()

	age	gender	split	final
0	33	M	3938.0	0.000008
1	32	M	3986.0	0.000008
2	31	M	4009.0	0.000008
3	38	M	3976.0	0.000008
4	31	M	3992.0	0.000008

marathon['frac'] = 1 - 2 * marathon['split'] / marathon["final"]
marathon.head()

	age	gender	split	final	frac
0	33	M	3938.0	7731.0	-0.018756
1	32	M	3986.0	7768.0	-0.026262
2	31	M	4009.0	7842.0	-0.022443
3	38	M	3976.0	8025.0	0.009097
4	31	M	3992.0	8039.0	0.006842

marathon['split_frac'] = np.where(marathon['frac']>0, 0, 1)
marathon.sample(10)

	age	gender	split	final	frac	split_frac
10199	43	M	6524.0	14839.0	0.120695	0
9853	36	W	6906.0	14724.0	0.061940	0
1385	39	M	5304.0	11185.0	0.051587	0
2923	51	M	5957.0	12162.0	0.020391	0
35513	42	M	9818.0	24291.0	0.191635	0
2812	32	M	5390.0	12090.0	0.108354	0
21482	60	M	7353.0	17853.0	0.176273	0
31825	59	M	8110.0	21184.0	0.234328	0
25189	38	W	7816.0	18852.0	0.170804	0
15959	22	W	7438.0	16395.0	0.092650	0