In [15]:
import pandas as pd
import numpy as np
# 载入数据
train = pd.read_csv('/Users/frank/Documents/workspace/kaggle/dataset/San_Francisco_Crime_Classification/train.csv', parse_dates = ['Dates'])
test = pd.read_csv('/Users/frank/Documents/workspace/kaggle/dataset/San_Francisco_Crime_Classification/test.csv', parse_dates = ['Dates'])
预览训练集
In [16]:
print train.head(10)
预览测试集合
In [17]:
print test.head(10)
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：
In [18]:
from sklearn import preprocessing
label = preprocessing.LabelEncoder()
label.fit([1, 2, 2, 6])
print label.transform([1, 1, 2, 6])
接下来我们对类别进行编码：
In [19]:
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：
In [20]:
pd.get_dummies(pd.Series(list('abca')))
Out[20]:
接下来对Dates、DayOfWeek、PdDistrict三个特征进行二值化处理：
In [21]:
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：
In [22]:
train_data = pd.concat([days, district, hour], axis=1)
train_data['crime'] = crime
针对测试集做同样的处理：
In [23]:
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：
In [24]:
print train_data.head(10)
print test_data.head(10)
分割训练集和验证集(70%训练,30%验证)准备建模：
In [35]:
from sklearn.cross_validation import train_test_split
training, validation = train_test_split(train_data, train_size=0.6)
In [37]:
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))
In [38]:
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))
在测试集上运行：
In [41]:
test_predicted = np.array(model.predict_proba(test_data[feature_list]))
保存结果：
In [44]:
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)