基于随机森林方法的缺失值填充

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from sklearn.datasets import load_boston  # 波士顿数据
from sklearn.impute import SimpleImputer  # 填充缺失值的类
from sklearn.ensemble import RandomForestRegressor  # 随机森林回归
from sklearn.model_selection import cross_val_score  # 交叉验证

dataset = load_boston()
dataset.data.shape  # 标签是连续型的值，用于回归分析
dataset.target[:5]  # 标签是连续的数值，连续型变量，用于回归问题

X_full, y_full = dataset.data, dataset.target
n_samples = X_full.shape[0]  # 506
n_features = X_full.shape[1] # 13

rng = np.random.RandomState(0)  # 确定随机种子
missing_rate = 0.5   # 缺失率是50%
# 计算缺失的样本总数；floor是向下取整
n_missing_samples = int(np.floor(n_samples * n_features * missing_rate))
n_missing_samples

# randint(下限,上限,n)：在上限和下限之间随机取出n个整数
missing_features = rng.randint(0, n_features, n_missing_samples) # (0,13,3289)
missing_samples = rng.randint(0, n_samples, n_missing_samples) # (0,506,3289)

# 采样了3289个数据，远远超过了样本量506，使用随机抽取的函数randint；
# 如果需要的数据量是小于样本量506，则需要使用randint.choice来抽样，保证抽取不重复的随机数
# missing_samples = rng.choice(n_samples, n_missing_samples, replace=False)

imp_mean = SimpleImputer(missing_values=np.nan, strategy="mean")  # 指定缺失值是什么和用什么填充
X_missing_mean = imp_mean.fit_transform(X_missing)  # fit + predict---->特殊接口transform

pd.DataFrame(X_missing_mean).isnull().sum() # X_missing_mean是一个ndaraay

imp_0 = SimpleImputer(missing_values=np.nan, strategy="constant", fill_value=0)   # 用0进行填充
X_missing_0 = imp_0.fit_transform(X_missing)

X_missing_reg = X_missing.copy()
# 找出缺失值从小到大对应的索引值
sortindex = np.argsort(X_missing_reg.isnull().sum(axis=0)).values

for i in sortindex:

    # 构建新的特征矩阵和新标签
    df = X_missing_reg   # 所有的操作都在df上进行，只是最后得到的填充值作用在X_missing_reg上面
    fillc = df.iloc[:, i]   # 某个需要填充的列，索引为i

    # 没有被选中填充（!=）的特征与原始标签的连接起来；df就是新特征矩阵
    df = pd.concat([df.iloc[:, df.columns != i], pd.DataFrame(y_full)], axis=1)


    # 新的特征矩阵df中，对含有缺失值的列，进行0的填补
    # 检查是否有0 pd.DataFrame(df_0).isnull().sum()
    df_0 = SimpleImputer(missing_values=np.nan, strategy='constant', fill_value=0).fit_transform(df)

    # 找出训练集和测试集
    ytrain = fillc[fillc.notnull()]  # 被选中填充的特征矩阵T中的非空值
    ytest = fillc[fillc.isnull()]  # 被选中填充的特征矩阵T中的空值
    Xtrain = df_0[ytrain.index, :]  # 新特征矩阵上，被选出来要填充的特征的非空值对应的记录
    Xtest = df_0[ytest.index, :]   # 空值对应的记录

    # 随机森林填充缺失值
    rfc = RandomForestRegressor(n_estimators=100)
    rfc = rfc.fit(Xtrain, ytrain)
    y_predict = rfc.predict(Xtest)  # predict接口预测得到的结果就是用来填充空值的那些值

    # 将填补好的特征返回到我们的原始特征矩阵中
    X_missing_reg.loc[X_missing_reg.iloc[:, i].isnull(), i] = y_predict

X = [X_full, X_missing_mean, X_missing_0, X_missing_reg]
mse = []

for x in X:
    estimator = RandomForestRegressor(random_state=0, n_estimators=100)
    scores = cross_val_score(estimator, x, y_full, scoring="neg_mean_squared_error", cv=5).mean()
    mse.append(scores * -1)

x_labels = ["Full Data",
           "Mean Imputation",
           "Zero Imputation",
           "Regressor Imputation"]

colors = ['r', 'g', 'b', 'orange']

plt.figure(figsize=(12,6))  # 生成画布
ax = plt.subplot(111)  # 添加子图

# 在上面的过程中已经求出了4种不同数据的MSE
for i in np.arange(len(mse)):  # arange(4)
    ax.barh(i, mse[i], color=colors[i],alpha=0.6,align='center')

ax.set_title("Imputation Techniques with Boston Data")  # 标题
ax.set_xlim(left=np.min(mse)*0.9,
           right=np.max(mse)*1.1)  # 横轴是mse的取值

ax.set_yticks(np.arange(len(mse)))
ax.set_label("MSE")
ax.set_yticklabels(x_labels)   # 用x_labels中的名字作为y轴的命名
plt.show()