异常检测-EllipticEnvelope与OneClassSVM的区别

1、EllipticEnvelope：the main result is that the empirical covariance estimate, as a non-robust one, is highly influenced by the heterogeneous structure of the observations. Although the robust covariance estimate is able to focus on the main mode of the data distribution, it sticks to the assumption that the data should be Gaussian distributed, yielding some biased estimation of the data structure, but yet accurate to some extent.

该模型最大的特点：数据为高斯分布；且鲁棒性不好；如果数据结构非常混杂，则效果不好。

2、One-Class SVM：The One-Class SVM does not assume any parametric form of the data distribution and can therefore model the complex shape of the data much better.

该模型最大的特点：对数据结构没有任何的假设，所以能更好的拟合数据。

案例一：

# Author: Virgile Fritsch <virgile.fritsch@inria.fr>
# License: BSD 3 clause

import numpy as np
from sklearn.covariance import EllipticEnvelope
from sklearn.svm import OneClassSVM
import matplotlib.pyplot as plt
import matplotlib.font_manager
from sklearn.datasets import load_wine

# Define "classifiers" to be used
classifiers = {
    "Empirical Covariance": EllipticEnvelope(support_fraction=1.,
                                             contamination=0.25),
    "Robust Covariance (Minimum Covariance Determinant)":
    EllipticEnvelope(contamination=0.25),
    "OCSVM": OneClassSVM(nu=0.25, gamma=0.35)}
colors = ['m', 'g', 'b']
legend1 = {}
legend2 = {}

# Get data
X1 = load_wine()['data'][:, [1, 2]]  # two clusters

# Learn a frontier for outlier detection with several classifiers
xx1, yy1 = np.meshgrid(np.linspace(0, 6, 500), np.linspace(1, 4.5, 500))
for i, (clf_name, clf) in enumerate(classifiers.items()):
    plt.figure(1)
    clf.fit(X1)
    Z1 = clf.decision_function(np.c_[xx1.ravel(), yy1.ravel()])
    Z1 = Z1.reshape(xx1.shape)
    legend1[clf_name] = plt.contour(
        xx1, yy1, Z1, levels=[0], linewidths=2, colors=colors[i])

legend1_values_list = list(legend1.values())
legend1_keys_list = list(legend1.keys())

# Plot the results (= shape of the data points cloud)
plt.figure(1)  # two clusters
plt.title("Outlier detection on a real data set (wine recognition)")
plt.scatter(X1[:, 0], X1[:, 1], color='black')
bbox_args = dict(boxstyle="round", fc="0.8")
arrow_args = dict(arrowstyle="->")
plt.annotate("outlying points", xy=(4, 2),
             xycoords="data", textcoords="data",
             xytext=(3, 1.25), bbox=bbox_args, arrowprops=arrow_args)
plt.xlim((xx1.min(), xx1.max()))
plt.ylim((yy1.min(), yy1.max()))
plt.legend((legend1_values_list[0].collections[0],
            legend1_values_list[1].collections[0],
            legend1_values_list[2].collections[0]),
           (legend1_keys_list[0], legend1_keys_list[1], legend1_keys_list[2]),
           loc="upper center",
           prop=matplotlib.font_manager.FontProperties(size=11))
plt.ylabel("ash")
plt.xlabel("malic_acid")

plt.show()

案例2 ：

# Get data
X2 = load_wine()['data'][:, [6, 9]]  # "banana"-shaped

# Learn a frontier for outlier detection with several classifiers
xx2, yy2 = np.meshgrid(np.linspace(-1, 5.5, 500), np.linspace(-2.5, 19, 500))
for i, (clf_name, clf) in enumerate(classifiers.items()):
    plt.figure(2)
    clf.fit(X2)
    Z2 = clf.decision_function(np.c_[xx2.ravel(), yy2.ravel()])
    Z2 = Z2.reshape(xx2.shape)
    legend2[clf_name] = plt.contour(
        xx2, yy2, Z2, levels=[0], linewidths=2, colors=colors[i])

legend2_values_list = list(legend2.values())
legend2_keys_list = list(legend2.keys())

# Plot the results (= shape of the data points cloud)
plt.figure(2)  # "banana" shape
plt.title("Outlier detection on a real data set (wine recognition)")
plt.scatter(X2[:, 0], X2[:, 1], color='black')
plt.xlim((xx2.min(), xx2.max()))
plt.ylim((yy2.min(), yy2.max()))
plt.legend((legend2_values_list[0].collections[0],
            legend2_values_list[1].collections[0],
            legend2_values_list[2].collections[0]),
           (legend2_keys_list[0], legend2_keys_list[1], legend2_keys_list[2]),
           loc="upper center",
           prop=matplotlib.font_manager.FontProperties(size=11))
plt.ylabel("color_intensity")
plt.xlabel("flavanoids")

plt.show()

结论：

1、从上述两个案例来看，OneClassSVM对数据分布拟合效果更好

2、如果数据服从高斯分布，那么第一种效果会更好，且泛化能力也强，所以对数据结构要有一定的了解