机器学习-数值特征

离散值处理

import pandas as pd
import numpy as np
1
2

vg_df = pd.read_csv('datasets/vgsales.csv', encoding = "ISO-8859-1")
vg_df[['Name', 'Platform', 'Year', 'Genre', 'Publisher']].iloc[1:7]
1
2

genres = np.unique(vg_df['Genre'])
genres
1
2

array(['Action', 'Adventure', 'Fighting', 'Misc', 'Platform', 'Puzzle',
       'Racing', 'Role-Playing', 'Shooter', 'Simulation', 'Sports',
       'Strategy'], dtype=object)
1
2
3

LabelEncoder

from sklearn.preprocessing import LabelEncoder

gle = LabelEncoder()
genre_labels = gle.fit_transform(vg_df['Genre'])
genre_mappings = {index: label for index, label in enumerate(gle.classes_)}
genre_mappings
1
2
3
4
5
6

{0: 'Action',
 1: 'Adventure',
 2: 'Fighting',
 3: 'Misc',
 4: 'Platform',
 5: 'Puzzle',
 6: 'Racing',
 7: 'Role-Playing',
 8: 'Shooter',
 9: 'Simulation',
 10: 'Sports',
 11: 'Strategy'}
1
2
3
4
5
6
7
8
9
10
11
12

vg_df['GenreLabel'] = genre_labels
vg_df[['Name', 'Platform', 'Year', 'Genre', 'GenreLabel']].iloc[1:7]
1
2

Map

poke_df = pd.read_csv('datasets/Pokemon.csv', encoding='utf-8')
poke_df = poke_df.sample(random_state=1, frac=1).reset_index(drop=True)

np.unique(poke_df['Generation'])
1
2
3
4

array(['Gen 1', 'Gen 2', 'Gen 3', 'Gen 4', 'Gen 5', 'Gen 6'], dtype=object)
1

gen_ord_map = {'Gen 1': 1, 'Gen 2': 2, 'Gen 3': 3, 
               'Gen 4': 4, 'Gen 5': 5, 'Gen 6': 6}

poke_df['GenerationLabel'] = poke_df['Generation'].map(gen_ord_map)
poke_df[['Name', 'Generation', 'GenerationLabel']].iloc[4:10]
1
2
3
4
5

One-hot Encoding

poke_df[['Name', 'Generation', 'Legendary']].iloc[4:10]
1

from sklearn.preprocessing import OneHotEncoder, LabelEncoder

# transform and map pokemon generations
gen_le = LabelEncoder()
gen_labels = gen_le.fit_transform(poke_df['Generation'])
poke_df['Gen_Label'] = gen_labels

# transform and map pokemon legendary status
leg_le = LabelEncoder()
leg_labels = leg_le.fit_transform(poke_df['Legendary'])
poke_df['Lgnd_Label'] = leg_labels

poke_df_sub = poke_df[['Name', 'Generation', 'Gen_Label', 'Legendary', 'Lgnd_Label']]
poke_df_sub.iloc[4:10]
1
2
3
4
5
6
7
8
9
10
11
12
13
14

# encode generation labels using one-hot encoding scheme
gen_ohe = OneHotEncoder()
gen_feature_arr = gen_ohe.fit_transform(poke_df[['Gen_Label']]).toarray()
gen_feature_labels = list(gen_le.classes_)
print (gen_feature_labels)
gen_features = pd.DataFrame(gen_feature_arr, columns=gen_feature_labels)

# encode legendary status labels using one-hot encoding scheme
leg_ohe = OneHotEncoder()
leg_feature_arr = leg_ohe.fit_transform(poke_df[['Lgnd_Label']]).toarray()
leg_feature_labels = ['Legendary_'+str(cls_label) for cls_label in leg_le.classes_]
print (leg_feature_labels)
leg_features = pd.DataFrame(leg_feature_arr, columns=leg_feature_labels)
1
2
3
4
5
6
7
8
9
10
11
12
13

['Gen 1', 'Gen 2', 'Gen 3', 'Gen 4', 'Gen 5', 'Gen 6']
['Legendary_False', 'Legendary_True']
1
2

poke_df_ohe = pd.concat([poke_df_sub, gen_features, leg_features], axis=1)
columns = sum([['Name', 'Generation', 'Gen_Label'],gen_feature_labels,
              ['Legendary', 'Lgnd_Label'],leg_feature_labels], [])
poke_df_ohe[columns].iloc[4:10]
1
2
3
4

Get Dummy

gen_dummy_features = pd.get_dummies(poke_df['Generation'], drop_first=True)
pd.concat([poke_df[['Name', 'Generation']], gen_dummy_features], axis=1).iloc[4:10]
1
2

gen_onehot_features = pd.get_dummies(poke_df['Generation'])
pd.concat([poke_df[['Name', 'Generation']], gen_onehot_features], axis=1).iloc[4:10]
1
2

import pandas as pd
import matplotlib.pyplot as plt
import matplotlib as mpl
import numpy as np
import scipy.stats as spstats

%matplotlib inline
mpl.style.reload_library()
mpl.style.use('classic')
mpl.rcParams['figure.facecolor'] = (1, 1, 1, 0)
mpl.rcParams['figure.figsize'] = [6.0, 4.0]
mpl.rcParams['figure.dpi'] = 100
1
2
3
4
5
6
7
8
9
10
11
12

poke_df = pd.read_csv('datasets/Pokemon.csv', encoding='utf-8')
poke_df.head()
1
2

poke_df[['HP', 'Attack', 'Defense']].head()
1

poke_df[['HP', 'Attack', 'Defense']].describe()
1

popsong_df = pd.read_csv('datasets/song_views.csv', encoding='utf-8')
popsong_df.head(10)
1
2

二值特征

watched = np.array(popsong_df['listen_count']) 
watched[watched >= 1] = 1
popsong_df['watched'] = watched
popsong_df.head(10)
1
2
3
4

from sklearn.preprocessing import Binarizer

bn = Binarizer(threshold=0.9)
pd_watched = bn.transform([popsong_df['listen_count']])[0]
popsong_df['pd_watched'] = pd_watched
popsong_df.head(11)
1
2
3
4
5
6

多项式特征

atk_def = poke_df[['Attack', 'Defense']]
atk_def.head()
1
2

from sklearn.preprocessing import PolynomialFeatures

pf = PolynomialFeatures(degree=2, interaction_only=False, include_bias=False)
res = pf.fit_transform(atk_def)
res
1
2
3
4
5

array([[    49.,     49.,   2401.,   2401.,   2401.],
       [    62.,     63.,   3844.,   3906.,   3969.],
       [    82.,     83.,   6724.,   6806.,   6889.],
       ..., 
       [   110.,     60.,  12100.,   6600.,   3600.],
       [   160.,     60.,  25600.,   9600.,   3600.],
       [   110.,    120.,  12100.,  13200.,  14400.]])
1
2
3
4
5
6
7

intr_features = pd.DataFrame(res, columns=['Attack', 'Defense', 'Attack^2', 'Attack x Defense', 'Defense^2'])
intr_features.head(5)
1
2

binning特征

fcc_survey_df = pd.read_csv('datasets/fcc_2016_coder_survey_subset.csv', encoding='utf-8')
fcc_survey_df[['ID.x', 'EmploymentField', 'Age', 'Income']].head()
1
2

fig, ax = plt.subplots()
fcc_survey_df['Age'].hist(color='#A9C5D3')
ax.set_title('Developer Age Histogram', fontsize=12)
ax.set_xlabel('Age', fontsize=12)
ax.set_ylabel('Frequency', fontsize=12)
1
2
3
4
5

Text(0,0.5,'Frequency')
1

Binning based on rounding

Age Range: Bin
---------------
 0 -  9  : 0
10 - 19  : 1
20 - 29  : 2
30 - 39  : 3
40 - 49  : 4
50 - 59  : 5
60 - 69  : 6
  ... and so on
1
2
3
4
5
6
7
8
9
10

fcc_survey_df['Age_bin_round'] = np.array(np.floor(np.array(fcc_survey_df['Age']) / 10.))
fcc_survey_df[['ID.x', 'Age', 'Age_bin_round']].iloc[1071:1076]
1
2

分位数切分

fcc_survey_df[['ID.x', 'Age', 'Income']].iloc[4:9]
1

fig, ax = plt.subplots()
fcc_survey_df['Income'].hist(bins=30, color='#A9C5D3')
ax.set_title('Developer Income Histogram', fontsize=12)
ax.set_xlabel('Developer Income', fontsize=12)
ax.set_ylabel('Frequency', fontsize=12)
1
2
3
4
5

Text(0,0.5,'Frequency')
1

quantile_list = [0, .25, .5, .75, 1.]
quantiles = fcc_survey_df['Income'].quantile(quantile_list)
quantiles
1
2
3

0.00      6000.0
0.25     20000.0
0.50     37000.0
0.75     60000.0
1.00    200000.0
Name: Income, dtype: float64
1
2
3
4
5
6

fig, ax = plt.subplots()
fcc_survey_df['Income'].hist(bins=30, color='#A9C5D3')

for quantile in quantiles:
    qvl = plt.axvline(quantile, color='r')
ax.legend([qvl], ['Quantiles'], fontsize=10)

ax.set_title('Developer Income Histogram with Quantiles', fontsize=12)
ax.set_xlabel('Developer Income', fontsize=12)
ax.set_ylabel('Frequency', fontsize=12)
1
2
3
4
5
6
7
8
9
10

Text(0,0.5,'Frequency')
1

quantile_labels = ['0-25Q', '25-50Q', '50-75Q', '75-100Q']
fcc_survey_df['Income_quantile_range'] = pd.qcut(fcc_survey_df['Income'], 
                                                 q=quantile_list)
fcc_survey_df['Income_quantile_label'] = pd.qcut(fcc_survey_df['Income'], 
                                                 q=quantile_list, labels=quantile_labels)
fcc_survey_df[['ID.x', 'Age', 'Income', 
               'Income_quantile_range', 'Income_quantile_label']].iloc[4:9]
1
2
3
4
5
6
7

对数变换 COX-BOX

fcc_survey_df['Income_log'] = np.log((1+ fcc_survey_df['Income']))
fcc_survey_df[['ID.x', 'Age', 'Income', 'Income_log']].iloc[4:9]
1
2

income_log_mean = np.round(np.mean(fcc_survey_df['Income_log']), 2)

fig, ax = plt.subplots()
fcc_survey_df['Income_log'].hist(bins=30, color='#A9C5D3')
plt.axvline(income_log_mean, color='r')
ax.set_title('Developer Income Histogram after Log Transform', fontsize=12)
ax.set_xlabel('Developer Income (log scale)', fontsize=12)
ax.set_ylabel('Frequency', fontsize=12)
ax.text(11.5, 450, r'$\mu$='+str(income_log_mean), fontsize=10)
1
2
3
4
5
6
7
8
9

Text(11.5,450,'$\\mu$=10.43')
1

日期相关特征

import datetime
import numpy as np
import pandas as pd
from dateutil.parser import parse
import pytz
1
2
3
4
5

time_stamps = ['2015-03-08 10:30:00.360000+00:00', '2017-07-13 15:45:05.755000-07:00',
               '2012-01-20 22:30:00.254000+05:30', '2016-12-25 00:30:00.000000+10:00']
df = pd.DataFrame(time_stamps, columns=['Time'])
df
1
2
3
4

ts_objs = np.array([pd.Timestamp(item) for item in np.array(df.Time)])
df['TS_obj'] = ts_objs
ts_objs
1
2
3

array([Timestamp('2015-03-08 10:30:00.360000+0000', tz='UTC'),
       Timestamp('2017-07-13 15:45:05.755000-0700', tz='pytz.FixedOffset(-420)'),
       Timestamp('2012-01-20 22:30:00.254000+0530', tz='pytz.FixedOffset(330)'),
       Timestamp('2016-12-25 00:30:00+1000', tz='pytz.FixedOffset(600)')], dtype=object)
1
2
3
4

df['Year'] = df['TS_obj'].apply(lambda d: d.year)
df['Month'] = df['TS_obj'].apply(lambda d: d.month)
df['Day'] = df['TS_obj'].apply(lambda d: d.day)
df['DayOfWeek'] = df['TS_obj'].apply(lambda d: d.dayofweek)
df['DayName'] = df['TS_obj'].apply(lambda d: d.weekday_name)
df['DayOfYear'] = df['TS_obj'].apply(lambda d: d.dayofyear)
df['WeekOfYear'] = df['TS_obj'].apply(lambda d: d.weekofyear)
df['Quarter'] = df['TS_obj'].apply(lambda d: d.quarter)

df[['Time', 'Year', 'Month', 'Day', 'Quarter', 
    'DayOfWeek', 'DayName', 'DayOfYear', 'WeekOfYear']]
1
2
3
4
5
6
7
8
9
10
11

时间相关特征

df['Hour'] = df['TS_obj'].apply(lambda d: d.hour)
df['Minute'] = df['TS_obj'].apply(lambda d: d.minute)
df['Second'] = df['TS_obj'].apply(lambda d: d.second)
df['MUsecond'] = df['TS_obj'].apply(lambda d: d.microsecond)   #毫秒
df['UTC_offset'] = df['TS_obj'].apply(lambda d: d.utcoffset()) #UTC时间位移

df[['Time', 'Hour', 'Minute', 'Second', 'MUsecond', 'UTC_offset']]
1
2
3
4
5
6
7

按照早晚切分时间

hour_bins = [-1, 5, 11, 16, 21, 23]
bin_names = ['Late Night', 'Morning', 'Afternoon', 'Evening', 'Night']
df['TimeOfDayBin'] = pd.cut(df['Hour'], 
                            bins=hour_bins, labels=bin_names)
df[['Time', 'Hour', 'TimeOfDayBin']]
1
2
3
4
5