特点：
- 离散特征
- 离散特征二值化处理
数据概览
import pandas as pd
import numpy as np
# 载入数据
train = pd.read_csv('~/kaggle/dataset/San_Francisco_Crime_Classification/train.csv', parse_dates = ['Dates'])
test = pd.read_csv('~/kaggle/dataset/San_Francisco_Crime_Classification/test.csv', parse_dates = ['Dates'])
预览训练集
print train.head(10)
Dates Category Descript \
0 2015-05-13 23:53:00 WARRANTS WARRANT ARREST
1 2015-05-13 23:53:00 OTHER OFFENSES TRAFFIC VIOLATION ARREST
2 2015-05-13 23:33:00 OTHER OFFENSES TRAFFIC VIOLATION ARREST
3 2015-05-13 23:30:00 LARCENY/THEFT GRAND THEFT FROM LOCKED AUTO
4 2015-05-13 23:30:00 LARCENY/THEFT GRAND THEFT FROM LOCKED AUTO
5 2015-05-13 23:30:00 LARCENY/THEFT GRAND THEFT FROM UNLOCKED AUTO
6 2015-05-13 23:30:00 VEHICLE THEFT STOLEN AUTOMOBILE
7 2015-05-13 23:30:00 VEHICLE THEFT STOLEN AUTOMOBILE
8 2015-05-13 23:00:00 LARCENY/THEFT GRAND THEFT FROM LOCKED AUTO
9 2015-05-13 23:00:00 LARCENY/THEFT GRAND THEFT FROM LOCKED AUTO
DayOfWeek PdDistrict Resolution Address \
0 Wednesday NORTHERN ARREST, BOOKED OAK ST / LAGUNA ST
1 Wednesday NORTHERN ARREST, BOOKED OAK ST / LAGUNA ST
2 Wednesday NORTHERN ARREST, BOOKED VANNESS AV / GREENWICH ST
3 Wednesday NORTHERN NONE 1500 Block of LOMBARD ST
4 Wednesday PARK NONE 100 Block of BRODERICK ST
5 Wednesday INGLESIDE NONE 0 Block of TEDDY AV
6 Wednesday INGLESIDE NONE AVALON AV / PERU AV
7 Wednesday BAYVIEW NONE KIRKWOOD AV / DONAHUE ST
8 Wednesday RICHMOND NONE 600 Block of 47TH AV
9 Wednesday CENTRAL NONE JEFFERSON ST / LEAVENWORTH ST
X Y
0 -122.425892 37.774599
1 -122.425892 37.774599
2 -122.424363 37.800414
3 -122.426995 37.800873
4 -122.438738 37.771541
5 -122.403252 37.713431
6 -122.423327 37.725138
7 -122.371274 37.727564
8 -122.508194 37.776601
9 -122.419088 37.807802
预览测试集合
print test.head(10)
Id Dates DayOfWeek PdDistrict Address \
0 0 2015-05-10 23:59:00 Sunday BAYVIEW 2000 Block of THOMAS AV
1 1 2015-05-10 23:51:00 Sunday BAYVIEW 3RD ST / REVERE AV
2 2 2015-05-10 23:50:00 Sunday NORTHERN 2000 Block of GOUGH ST
3 3 2015-05-10 23:45:00 Sunday INGLESIDE 4700 Block of MISSION ST
4 4 2015-05-10 23:45:00 Sunday INGLESIDE 4700 Block of MISSION ST
5 5 2015-05-10 23:40:00 Sunday TARAVAL BROAD ST / CAPITOL AV
6 6 2015-05-10 23:30:00 Sunday INGLESIDE 100 Block of CHENERY ST
7 7 2015-05-10 23:30:00 Sunday INGLESIDE 200 Block of BANKS ST
8 8 2015-05-10 23:10:00 Sunday MISSION 2900 Block of 16TH ST
9 9 2015-05-10 23:10:00 Sunday CENTRAL TAYLOR ST / GREEN ST
X Y
0 -122.399588 37.735051
1 -122.391523 37.732432
2 -122.426002 37.792212
3 -122.437394 37.721412
4 -122.437394 37.721412
5 -122.459024 37.713172
6 -122.425616 37.739351
7 -122.412652 37.739750
8 -122.418700 37.765165
9 -122.413935 37.798886
我们看到训练集和测试集都有Dates、DayOfWeek、PdDistrict三个特征，我们先从这三个特征入手。训练集中的Category是我们的预测目标，我们先对其进行编码，这里用到sklearn的LabelEncoder()，示例如下：
from sklearn import preprocessing
label = preprocessing.LabelEncoder()
label.fit([1, 2, 2, 6])
print label.transform([1, 1, 2, 6])
[0 0 1 2]
接下来我们对类别进行编码：
crime = label.fit_transform(train.Category)
对于离散化的特征，有一种常用的特征处理方式是二值化处理，pandas中有get_dummies()函数，函数示例如下：
pd.get_dummies(pd.Series(list('abca')))
| a | b | c | |
|---|---|---|---|
| 0 | 1.0 | 0.0 | 0.0 |
| 1 | 0.0 | 1.0 | 0.0 |
| 2 | 0.0 | 0.0 | 1.0 |
| 3 | 1.0 | 0.0 | 0.0 |
接下来对Dates、DayOfWeek、PdDistrict三个特征进行二值化处理：
days = pd.get_dummies(train.DayOfWeek)
district = pd.get_dummies(train.PdDistrict)
hour = pd.get_dummies(train.Dates.dt.hour)
接下来重新组合训练集，并把类别附加上：
train_data = pd.concat([days, district, hour], axis=1)
train_data['crime'] = crime
针对测试集做同样的处理：
days = pd.get_dummies(test.DayOfWeek)
district = pd.get_dummies(test.PdDistrict)
hour = pd.get_dummies(test.Dates.dt.hour)
test_data = pd.concat([days, district, hour], axis=1)
预览新的训练集和测试集：
print train_data.head(10)
print test_data.head(10)
Friday Monday Saturday Sunday Thursday Tuesday Wednesday BAYVIEW \
0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
1 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
2 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
3 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
4 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
5 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
6 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
7 0.0 0.0 0.0 0.0 0.0 0.0 1.0 1.0
8 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
9 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0
CENTRAL INGLESIDE ... 15 16 17 18 19 20 21 22 23 \
0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
1 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
2 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
3 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
4 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
5 0.0 1.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
6 0.0 1.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
7 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
8 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
9 1.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
crime
0 37
1 21
2 21
3 16
4 16
5 16
6 36
7 36
8 16
9 16
[10 rows x 42 columns]
Friday Monday Saturday Sunday Thursday Tuesday Wednesday BAYVIEW \
0 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
1 0.0 0.0 0.0 1.0 0.0 0.0 0.0 1.0
2 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0
3 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0
4 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0
5 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0
6 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0
7 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0
8 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0
9 0.0 0.0 0.0 1.0 0.0 0.0 0.0 0.0
CENTRAL INGLESIDE ... 14 15 16 17 18 19 20 21 22 23
0 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
1 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
2 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
3 0.0 1.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
4 0.0 1.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
5 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
6 0.0 1.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
7 0.0 1.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
8 0.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
9 1.0 0.0 ... 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0
[10 rows x 41 columns]
分割训练集和验证集(70%训练,30%验证)准备建模：
from sklearn.cross_validation import train_test_split
training, validation = train_test_split(train_data, train_size=0.6)
贝叶斯训练
from sklearn.metrics import log_loss
from sklearn.naive_bayes import BernoulliNB
model = BernoulliNB()
feature_list = training.columns.tolist()
feature_list = feature_list[:len(feature_list) - 1]
print '选取的特征列：', feature_list
model.fit(training[feature_list], training['crime'])
predicted = np.array(model.predict_proba(validation[feature_list]))
print "朴素贝叶斯log损失为 %f" % (log_loss(validation['crime'], predicted))
选取的特征列： ['Friday', 'Monday', 'Saturday', 'Sunday', 'Thursday', 'Tuesday', 'Wednesday', 'BAYVIEW', 'CENTRAL', 'INGLESIDE', 'MISSION', 'NORTHERN', 'PARK', 'RICHMOND', 'SOUTHERN', 'TARAVAL', 'TENDERLOIN', 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23]
朴素贝叶斯log损失为 2.581561
逻辑回归
from sklearn.linear_model import LogisticRegression
model = LogisticRegression(C=0.1)
model.fit(training[feature_list], training['crime'])
predicted = np.array(model.predict_proba(validation[feature_list]))
print "逻辑回归log损失为 %f" %(log_loss(validation['crime'], predicted))
逻辑回归log损失为 2.580102
在测试集上运行：
test_predicted = np.array(model.predict_proba(test_data[feature_list]))
保存结果：
col_names = np.sort(train['Category'].unique())
print col_names
result = pd.DataFrame(data=test_predicted, columns=col_names)
result['Id'] = test['Id'].astype(int)
result.to_csv('output.csv', index=False)
['ARSON' 'ASSAULT' 'BAD CHECKS' 'BRIBERY' 'BURGLARY' 'DISORDERLY CONDUCT'
'DRIVING UNDER THE INFLUENCE' 'DRUG/NARCOTIC' 'DRUNKENNESS' 'EMBEZZLEMENT'
'EXTORTION' 'FAMILY OFFENSES' 'FORGERY/COUNTERFEITING' 'FRAUD' 'GAMBLING'
'KIDNAPPING' 'LARCENY/THEFT' 'LIQUOR LAWS' 'LOITERING' 'MISSING PERSON'
'NON-CRIMINAL' 'OTHER OFFENSES' 'PORNOGRAPHY/OBSCENE MAT' 'PROSTITUTION'
'RECOVERED VEHICLE' 'ROBBERY' 'RUNAWAY' 'SECONDARY CODES'
'SEX OFFENSES FORCIBLE' 'SEX OFFENSES NON FORCIBLE' 'STOLEN PROPERTY'
'SUICIDE' 'SUSPICIOUS OCC' 'TREA' 'TRESPASS' 'VANDALISM' 'VEHICLE THEFT'
'WARRANTS' 'WEAPON LAWS']