Python实现异常值自动检测的案例分享

import numpy as np
from scipy.stats import zscore

# 生成随机数据
data = np.random.randn(100)

# 计算Z-Score
z_scores = zscore(data)

# 定义阈值，判断是否为异常值
threshold = 2.5
outliers = np.where(np.abs(z_scores) > threshold)[0]

print("Z-Score Outliers:", outliers)

from sklearn.ensemble import IsolationForest

# 生成随机数据
data = np.random.randn(100, 1)

# 创建Isolation Forest模型
model = IsolationForest(contamination=0.1)  # contamination参数表示异常值的比例

# 训练模型并预测异常值
outliers = model.fit_predict(data)

print("Isolation Forest Outliers:", np.where(outliers == -1)[0])

import matplotlib.pyplot as plt
import seaborn as sns

# 生成带有异常值的随机数据
data = np.concatenate([np.random.randn(80), np.random.randn(20) * 5 + 10])

# 绘制直方图
plt.figure(figsize=(10, 6))
sns.histplot(data, kde=True)
plt.title("Histogram with Outliers")
plt.show()

# 绘制箱线图
plt.figure(figsize=(10, 6))
sns.boxplot(x=data)
plt.title("Boxplot with Outliers")
plt.show()

# 假设我们将异常值替换为中位数
median_value = np.median(data)
data_no_outliers = np.where(np.abs(z_scores) > threshold, median_value, data)

# 绘制处理后的数据
plt.figure(figsize=(10, 6))
sns.histplot(data_no_outliers, kde=True)
plt.title("Histogram without Outliers")
plt.show()

from sklearn.cluster import DBSCAN
from sklearn.preprocessing import StandardScaler

# 生成带有异常值的随机数据
data = np.concatenate([np.random.randn(80), np.random.randn(20) * 5 + 10])

# 使用DBSCAN进行聚类
data = data.reshape(-1, 1)  # DBSCAN要求输入是二维数组
data_scaled = StandardScaler().fit_transform(data)
dbscan = DBSCAN(eps=0.5, min_samples=5)
clusters = dbscan.fit_predict(data_scaled)

# 找出标签为-1的异常点
outliers = np.where(clusters == -1)[0]

print("DBSCAN Outliers:", outliers)

import seaborn as sns

# 加载鸢尾花数据集
iris = sns.load_dataset("iris")

# 选择一个特征进行演示（这里选择花萼长度）
feature = "sepal_length"
data_iris = iris[feature]

# 计算Z-Score
z_scores_iris = zscore(data_iris)

# 定义阈值，判断是否为异常值
threshold_iris = 2.5
outliers_iris = np.where(np.abs(z_scores_iris) > threshold_iris)[0]

print("Z-Score Outliers in Iris Dataset:", outliers_iris)

# 绘制箱线图
plt.figure(figsize=(10, 6))
sns.boxplot(x=data_iris)
plt.title("Boxplot for Sepal Length in Iris Dataset")
plt.show()

# 假设我们将异常值替换为中位数
median_value_iris = np.median(data_iris)
data_iris_no_outliers = np.where(np.abs(z_scores_iris) > threshold_iris, median_value_iris, data_iris)

# 绘制处理后的数据
plt.figure(figsize=(10, 6))
sns.histplot(data_iris_no_outliers, kde=True)
plt.title("Histogram for Sepal Length without Outliers")
plt.show()