Xgboost筛选特征重要性

基本思想

根据结构分数的增益情况计算出来选择哪个特征的哪个分割点，某个特征的重要性，就是它在所有树中出现的次数之和。

使用代码

import pandas as pd
import xgboost as xgb
import operator
from matplotlib import pylab as plt

def ceate_feature_map(features):
    outfile = open('xgb.fmap', 'w')
    i = 0
    for feat in features:
        outfile.write('{0}\t{1}\tq\n'.format(i, feat))
        i = i + 1

    outfile.close()

def get_data():
    train = pd.read_csv("../input/train.csv")

    features = list(train.columns[2:])

    y_train = train.Hazard

    for feat in train.select_dtypes(include=['object']).columns:
        m = train.groupby([feat])['Hazard'].mean()
        train[feat].replace(m,inplace=True)

    x_train = train[features]

    return features, x_train, y_train

def get_data2():
    from sklearn.datasets import load_iris
    #获取数据
    iris = load_iris()
    x_train=pd.DataFrame(iris.data)
    features=["sepal_length","sepal_width","petal_length","petal_width"]
    x_train.columns=features
    y_train=pd.DataFrame(iris.target)
    return features, x_train, y_train

#features, x_train, y_train = get_data()
features, x_train, y_train = get_data2()
ceate_feature_map(features)

xgb_params = {"objective": "reg:linear", "eta": 0.01, "max_depth": 8, "seed": 42, "silent": 1}
num_rounds = 1000

dtrain = xgb.DMatrix(x_train, label=y_train)
xgb_model = xgb.train(xgb_params, dtrain, num_rounds)

importance = xgb_model.get_fscore(fmap='xgb.fmap')
importance = sorted(importance.items(), key=operator.itemgetter(1))

df = pd.DataFrame(importance, columns=['feature', 'fscore'])
df['fscore'] = df['fscore'] / df['fscore'].sum()

plt.figure()
df.plot()
df.plot(kind='barh', x='feature', y='fscore', legend=False, figsize=(16, 10))
plt.title('XGBoost Feature Importance')
plt.xlabel('relative importance')
plt.gcf().savefig('feature_importance_xgb.png')

import pandas as pd
import matplotlib.pylab as plt
feat_imp = pd.Series(clf.booster().get_fscore()).sort_values(ascending=False)
#新版需要转换成dict or list 
#feat_imp = pd.Series(dict(clf.get_booster().get_fscore())).sort_values(ascending=False)
#plt.bar(feat_imp.index, feat_imp)  

feat_imp.plot(kind='bar', title='Feature Importances')
plt.ylabel('Feature Importance Score')
plt.show()