跟着教程学习了一段时间数据分析,越学感觉坑越多。于是花了一个星期仔细看了下《利用Python进行数据分析》。写在这里主要是记录下,方便自己查看。
scikit-learn是一个被广泛使用的python机器学习工具包。里面包含了很多监督式学习和非监督式学习的模型,可以实现分类,聚类,预测等任务。
虽然scikit-learn并没有和pandas深度整合,但在训练模型之前,pandas在数据清洗阶段能起很大作用。
译者:构建的机器学习模型的一个常见流程是,用pandas对数据进行查看和清洗,然后把处理过的数据喂给scikit-learn中的模型进行训练。
这里用一个经典的kaggle比赛数据集来做例子,泰坦尼克生还者数据集。加载训练集和测试集:
import numpy as np
import pandas as pd
train = pd.read_csv('../datasets/titanic/train.csv')
test = pd.read_csv('../datasets/titanic/test.csv')
train.head()
PassengerId | Survived | Pclass | Name | Sex | Age | SibSp | Parch | Ticket | Fare | Cabin | Embarked | |
---|---|---|---|---|---|---|---|---|---|---|---|---|
0 | 1 | 0 | 3 | Braund, Mr. Owen Harris | male | 22.0 | 1 | 0 | A/5 21171 | 7.2500 | NaN | S |
1 | 2 | 1 | 1 | Cumings, Mrs. John Bradley (Florence Briggs Th… | female | 38.0 | 1 | 0 | PC 17599 | 71.2833 | C85 | C |
2 | 3 | 1 | 3 | Heikkinen, Miss. Laina | female | 26.0 | 0 | 0 | STON/O2. 3101282 | 7.9250 | NaN | S |
3 | 4 | 1 | 1 | Futrelle, Mrs. Jacques Heath (Lily May Peel) | female | 35.0 | 1 | 0 | 113803 | 53.1000 | C123 | S |
4 | 5 | 0 | 3 | Allen, Mr. William Henry | male | 35.0 | 0 | 0 | 373450 | 8.0500 | NaN | S |
statsmodels和scikit-learn通常不能应付缺失值,所以我们先检查一下哪些列有缺失值:
train.isnull().sum()
PassengerId 0
Survived 0
Pclass 0
Name 0
Sex 0
Age 177
SibSp 0
Parch 0
Ticket 0
Fare 0
Cabin 687
Embarked 2
dtype: int64
test.isnull().sum()
PassengerId 0
Pclass 0
Name 0
Sex 0
Age 86
SibSp 0
Parch 0
Ticket 0
Fare 1
Cabin 327
Embarked 0
dtype: int64
对于这样的数据集,通常的任务是预测一个乘客最后是否生还。在训练集上训练模型,在测试集上验证效果。
上面的Age这一列有缺失值,这里我们简单的用中位数来代替缺失值:
impute_value = train['Age'].median()
train['Age'] = train['Age'].fillna(impute_value)
test['Age'] = test['Age'].fillna(impute_value)
对于Sex列,我们将其变为IsFemale,用整数来表示性别:
train['IsFemale'] = (train['Sex'] == 'female').astype(int)
test['IsFemale'] = (test['Sex'] == 'female').astype(int)
train.head()
PassengerId | Survived | Pclass | Name | Sex | Age | SibSp | Parch | Ticket | Fare | Cabin | Embarked | IsFemale | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
0 | 1 | 0 | 3 | Braund, Mr. Owen Harris | male | 22.0 | 1 | 0 | A/5 21171 | 7.2500 | NaN | S | 0 |
1 | 2 | 1 | 1 | Cumings, Mrs. John Bradley (Florence Briggs Th… | female | 38.0 | 1 | 0 | PC 17599 | 71.2833 | C85 | C | 1 |
2 | 3 | 1 | 3 | Heikkinen, Miss. Laina | female | 26.0 | 0 | 0 | STON/O2. 3101282 | 7.9250 | NaN | S | 1 |
3 | 4 | 1 | 1 | Futrelle, Mrs. Jacques Heath (Lily May Peel) | female | 35.0 | 1 | 0 | 113803 | 53.1000 | C123 | S | 1 |
4 | 5 | 0 | 3 | Allen, Mr. William Henry | male | 35.0 | 0 | 0 | 373450 | 8.0500 | NaN | S | 0 |
接下来决定一些模型参数并创建numpy数组:
predictors = ['Pclass', 'IsFemale', 'Age']
X_train = train[predictors].values
X_test = test[predictors].values
y_train = train['Survived'].values
X_train[:5]
array([[ 3., 0., 22.],
[ 1., 1., 38.],
[ 3., 1., 26.],
[ 1., 1., 35.],
[ 3., 0., 35.]])
y_train[:5]
array([0, 1, 1, 1, 0], dtype=int64)
这里我们用逻辑回归模型(LogisticRegression):
from sklearn.linear_model import LogisticRegression
model = LogisticRegression()
然后是fit方法来拟合模型:
model.fit(X_train, y_train)
LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,
intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,
penalty='l2', random_state=None, solver='liblinear', tol=0.0001,
verbose=0, warm_start=False)
在测试集上进行预测,使用model.predict:
y_predict = model.predict(X_test)
y_predict[:10]
array([0, 0, 0, 0, 1, 0, 1, 0, 1, 0], dtype=int64)
如果我们有测试集的真是结果的话,可以用来计算准确率或其他一些指标:
(y_true == y_predcit).mean()
实际过程中,训练模型的时候,经常用到交叉验证(cross-validation),用于调参,防止过拟合。这样得到的预测效果会更好,健壮性更强。
交叉验证是把训练集分为几份,每一份上又取出一部分作为测试样本,这些被取出来的测试样本不被用于训练,但我们可以在这些测试样本上验证当前模型的准确率或均方误差(mean squared error),而且还可以在模型参数上进行网格搜索(grid search)。一些模型,比如逻辑回归,自带一个有交叉验证的类。LogisticRegressionCV类可以用于模型调参,使用的时候需要指定正则化项C,来控制网格搜索的程度:
from sklearn.linear_model import LogisticRegressionCV
model_cv = LogisticRegressionCV(10)
model_cv.fit(X_train, y_train)
LogisticRegressionCV(Cs=10, class_weight=None, cv=None, dual=False,
fit_intercept=True, intercept_scaling=1.0, max_iter=100,
multi_class='ovr', n_jobs=1, penalty='l2', random_state=None,
refit=True, scoring=None, solver='lbfgs', tol=0.0001, verbose=0)
如果想要自己来做交叉验证的话,可以使用cross_val_score函数,可以用于数据切分。比如,把整个训练集分为4个不重叠的部分:
from sklearn.model_selection import cross_val_score
model = LogisticRegression(C=10)
model
LogisticRegression(C=10, class_weight=None, dual=False, fit_intercept=True,
intercept_scaling=1, max_iter=100, multi_class='ovr', n_jobs=1,
penalty='l2', random_state=None, solver='liblinear', tol=0.0001,
verbose=0, warm_start=False)
scores = cross_val_score(model, X_train, y_train, cv=4)
scores
array([0.77232143, 0.80269058, 0.77027027, 0.78828829])
默认的评价指标每个模型是不一样的,但是可以自己指定评价函数。交差验证的训练时间较长,但通常能得到更好的模型效果。