python降维分析_Python数据降维

一些数据降维的特征提取算法，先导入包和数据：

import numpy as np

import pandas as pd

import matplotlib.pyplot as plt

from mpl_toolkits.mplot3d import Axes3D

from sklearn.decomposition import PCA, KernelPCA

from sklearn.manifold import Isomap, MDS, TSNE

from sklearn.preprocessing import StandardScaler

## 对酒的特征数据进行标准化

wine_x,wine_y = load_wine(return_X_y=True)

wine_x = StandardScaler().fit_transform(wine_x)

主成分分析

## 使用主成分分析对酒数据集进行降维

pca = PCA(n_components = 13,random_state = 123)

pca.fit(wine_x)

## 使用数据的前3个主成分较合适

pca_wine_x = pca.transform(wine_x)[:,0:3]

print(pca_wine_x.shape)

## 在3D空间中可视化主成分分析后的数据空间分布

colors = ["red","blue","green"]

shapes = ["o","s","*"]

fig = plt.figure(figsize=(10,6))

## 将坐标系设置为3D

ax1 = fig.add_subplot(111, projection="3d")

for ii,y in enumerate(wine_y):

ax1.scatter(pca_wine_x[ii,0],pca_wine_x[ii,1],pca_wine_x[ii,2],

s = 40,c = colors[y],marker = shapes[y])

ax1.set_xlabel("主成分1",rotation=20)

ax1.set_ylabel("主成分2",rotation=-20)

ax1.set_zlabel("主成分3",rotation=90)

ax1.azim = 225

ax1.set_title("主成分特征空间可视化")

plt.show()

核主成分分析

## 使用核主成分分析获取数据的主成分

kpca = KernelPCA(n_components = 13,kernel = "rbf", ## 核函数为rbf核

gamma = 0.2,random_state = 123)

kpca.fit(wine_x)

## 获取前3个核主成分

kpca_wine_x = kpca.transform(wine_x)[:,0:3]

print(kpca_wine_x.shape)

## 在3D空间中可视化主成分分析后的数据空间分布

colors = ["red","blue","green"]

shapes = ["o","s","*"]

fig = plt.figure(figsize=(10,6))

## 将坐标系设置为3D

ax1 = fig.add_subplot(111, projection="3d")

for ii,y in enumerate(wine_y):

ax1.scatter(kpca_wine_x[ii,0],kpca_wine_x[ii,1],kpca_wine_x[ii,2],

s = 40,c = colors[y],marker = shapes[y])

ax1.set_xlabel("核主成分1",rotation=20)

ax1.set_ylabel("核主成分2",rotation=-20)

ax1.set_zlabel("核主成分3",rotation=90)

ax1.azim = 225

ax1.set_title("核主成分特征空间可视化")

plt.show()

流形学习

## 流行学习进行数据的非线性降维

isomap = Isomap(n_neighbors = 7,## 每个点考虑的近邻数量

n_components = 3) ## 降维到3维空间中

## 获取降维后的数据

isomap_wine_x = isomap.fit_transform(wine_x)

print(isomap_wine_x.shape)

## 在3D空间中可视化流行降维后的数据空间分布

colors = ["red","blue","green"]

shapes = ["o","s","*"]

fig = plt.figure(figsize=(10,6))

## 将坐标系设置为3D

ax1 = fig.add_subplot(111, projection="3d")

for ii,y in enumerate(wine_y):

ax1.scatter(isomap_wine_x[ii,0],isomap_wine_x[ii,1],isomap_wine_x[ii,2],

s = 40,c = colors[y],marker = shapes[y])

ax1.set_xlabel("特征1",rotation=20)

ax1.set_ylabel("特征2",rotation=-20)

ax1.set_zlabel("特征3",rotation=90)

ax1.azim = 225

ax1.set_title("Isomap降维可视化")

plt.show()

TSNE

## TSNE进行数据的降维,降维到3维空间中

tsne = TSNE(n_components = 3,perplexity =25,

early_exaggeration =3,random_state=123)

## 获取降维后的数据

tsne_wine_x = tsne.fit_transform(wine_x)

print(tsne_wine_x.shape)

## 在3D空间中可视化流行降维后的数据空间分布

colors = ["red","blue","green"]

shapes = ["o","s","*"]

fig = plt.figure(figsize=(10,6))

## 将坐标系设置为3D

ax1 = fig.add_subplot(111, projection="3d")

for ii,y in enumerate(wine_y):

ax1.scatter(tsne_wine_x[ii,0],tsne_wine_x[ii,1],tsne_wine_x[ii,2],

s = 40,c = colors[y],marker = shapes[y])

ax1.set_xlabel("特征1",rotation=20)

ax1.set_ylabel("特征2",rotation=-20)

ax1.set_zlabel("特征3",rotation=90)

ax1.azim = 225

ax1.set_title("Isomap降维可视化")

plt.show()

多维尺度分析

## MDS进行数据的降维,降维到3维空间中

mds = MDS(n_components = 3,dissimilarity = "euclidean",random_state=123)

## 获取降维后的数据

mds_wine_x = mds.fit_transform(wine_x)

print(mds_wine_x.shape)

## 在3D空间中可视化流行降维后的数据空间分布

colors = ["red","blue","green"]

shapes = ["o","s","*"]

fig = plt.figure(figsize=(10,6))

## 将坐标系设置为3D

ax1 = fig.add_subplot(111, projection="3d")

for ii,y in enumerate(wine_y):

ax1.scatter(mds_wine_x[ii,0],mds_wine_x[ii,1],mds_wine_x[ii,2],

s = 40,c = colors[y],marker = shapes[y])

ax1.set_xlabel("特征1",rotation=20)

ax1.set_ylabel("特征2",rotation=-20)

ax1.set_zlabel("特征3",rotation=90)

ax1.azim = 225

ax1.set_title("MDS降维可视化")

plt.show()