孤立森林是一个经典的异常检测算法，能处理大规模的多维数据。

​ 孤立森林认为异常样本通常few and different：相比正常样本，它们数量上比较少，特征值差异比较大。因此，异常样本更容易被孤立。孤立森林通过构建二叉树的方法孤立每一个异常样本——因为异常样本容易被孤立的特征，异常样本更靠近根节点，正常样本则在更深处。构建的这棵树被称为孤立树（Isolation Tree, iTree)。iTrees的集合即iForest。anomalies are those instances which have short average path lengths on the iTrees。

from sklearn.ensemble import IsolationForest

IsolationForest不能将检测为异常值的某个数据设为缺失值只能删除

代码

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20

import numpy as np
import pandas as pd
from sklearn.feature_selection import VarianceThreshold
from sklearn.model_selection import GridSearchCV
from sklearn.feature_selection import SelectKBest
from sklearn.feature_selection import chi2
from sklearn.feature_selection import mutual_info_classif as MIC

from xgboost import XGBClassifier
from sklearn.metrics import f1_score
import random
import warnings
warnings.filterwarnings("ignore")  # 忽略警告信息
data_train = pd.read_csv("../原始数据集/train_10000.csv")
data_test = pd.read_csv("../原始数据集/validate_1000.csv")
#获取数据
X_val_df = data_test.iloc[:, 1:-1]
y_val_df = data_test.iloc[:, -1:]
X_train_df = data_train.iloc[:, 1:-1]
y_train_df = data_train.iloc[:, -1:]

1
2

df = X_train_df.fillna(X_train_df.median())
X_train_df = X_train_df.fillna(X_train_df.median())

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

from sklearn.ensemble import IsolationForest
import numpy as np

# 创建示例数据
X = np.random.rand(200, 2)

# 创建孤立森林模型
clf = IsolationForest(contamination=0.05, n_estimators=200, random_state=42)  # contamination参数用于指定异常值的比例

# 拟合模型并预测异常值
clf.fit(df.values)
y_pred = clf.predict(df.values)

# 打印异常值的预测结果
print("异常值预测结果：")
print(y_pred)

异常值预测结果：
[1 1 1 ... 1 1 1]

1
2

df = pd.DataFrame(y_pred)
df.to_csv('output.csv', index=False)

1
2
3
4
5
6
7
8
9

bool_array = y_pred == -1
# bool_l = bool_l.tolist()
# bool_l
# 根据布尔数组剔除样本
df_filtered = df[~bool_array]

# 打印剔除样本后的DataFrame
df_filtered

1

y_train_flt = y_train_df[~bool_array]

1
2
3
4

X_train_flt = df_filtered.values
X_train = X_train_df.values
y_train = y_train_df.values
y_train_flt = y_train_flt.values

1

X_train.shape, X_train_flt.shape, y_train.shape, y_train_flt.shape

((10000, 107), (9500, 107), (10000, 1), (9500, 1))

1
2
3
4

clf1 = XGBClassifier()
clf2 = XGBClassifier()
clf1.fit(X_train, y_train)
clf2.fit(X_train_flt, y_train_flt)

XGBClassifier(base_score=0.5, booster='gbtree', callbacks=None,
              colsample_bylevel=1, colsample_bynode=1, colsample_bytree=1,
              early_stopping_rounds=None, enable_categorical=False,
              eval_metric=None, gamma=0, gpu_id=-1, grow_policy='depthwise',
              importance_type=None, interaction_constraints='',
              learning_rate=0.300000012, max_bin=256, max_cat_to_onehot=4,
              max_delta_step=0, max_depth=6, max_leaves=0, min_child_weight=1,
              missing=nan, monotone_constraints='()', n_estimators=100,
              n_jobs=0, num_parallel_tree=1, objective='multi:softprob',
              predictor='auto', random_state=0, reg_alpha=0, ...)

1
2
3
4
5

from sklearn.metrics import classification_report
X_val = X_val_df.values
y_val = y_val_df.values
print(classification_report(y_val, clf1.predict(X_val)))
print(classification_report(y_val, clf2.predict(X_val)))

              precision    recall  f1-score   support

           0       0.58      0.74      0.65       176
           1       0.48      0.67      0.56       166
           2       0.64      0.40      0.49       171
           3       0.96      0.89      0.92       169
           4       0.89      0.74      0.81       156
           5       1.00      0.93      0.96       162

    accuracy                           0.73      1000
   macro avg       0.76      0.73      0.73      1000
weighted avg       0.75      0.73      0.73      1000

              precision    recall  f1-score   support

           0       0.78      0.64      0.70       176
           1       0.47      0.76      0.58       166
           2       0.68      0.50      0.58       171
           3       0.94      0.90      0.92       169
           4       0.90      0.88      0.89       156
           5       0.99      0.91      0.95       162

    accuracy                           0.76      1000
   macro avg       0.79      0.76      0.77      1000
weighted avg       0.79      0.76      0.77      1000

​