一些資料降維的特徵提取演算法，先導入包和資料：
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
## 對酒的特徵資料進行標準化
wine_x,wine_y = load_wine(return_x_y=true)
wine_x = standardscaler().fit_transform(wine_x)
主成分分析
## 使用主成分分析對酒資料集進行降維
pca = pca(n_components = 13,random_state = 123)
pca.fit(wine_x)
## 使用資料的前3個主成分較合適
pca_wine_x = pca.transform(wine_x)[:,0:3]
print(pca_wine_x.shape)
## 在3d空間中視覺化主成分分析後的資料空間分布
colors = ["red","blue","green"]
shapes = ["o","s","*"]
fig = plt.figure(figsize=(10,6))
## 將座標系設定為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("主成分特徵空間視覺化")
plt.show()
核主成分分析
## 使用核主成分分析獲取資料的主成分
kpca = kernelpca(n_components = 13,kernel = "rbf", ## 核函式為rbf核
gamma = 0.2,random_state = 123)
kpca.fit(wine_x)
## 獲取前3個核主成分
kpca_wine_x = kpca.transform(wine_x)[:,0:3]
print(kpca_wine_x.shape)
## 在3d空間中視覺化主成分分析後的資料空間分布
colors = ["red","blue","green"]
shapes = ["o","s","*"]
fig = plt.figure(figsize=(10,6))
## 將座標系設定為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("核主成分特徵空間視覺化")
plt.show()
流形學習
## 流行學習進行資料的非線性降維
isomap = isomap(n_neighbors = 7,## 每個點考慮的近鄰數量
n_components = 3) ## 降維到3維空間中
## 獲取降維後的資料
isomap_wine_x = isomap.fit_transform(wine_x)
print(isomap_wine_x.shape)
## 在3d空間中視覺化流行降維後的資料空間分布
colors = ["red","blue","green"]
shapes = ["o","s","*"]
fig = plt.figure(figsize=(10,6))
## 將座標系設定為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("特徵1",rotation=20)
ax1.set_ylabel("特徵2",rotation=-20)
ax1.set_zlabel("特徵3",rotation=90)
ax1.azim = 225
ax1.set_title("isomap降維視覺化")
plt.show()
tsne
## tsne進行資料的降維,降維到3維空間中
tsne = tsne(n_components = 3,perplexity =25,
early_exaggeration =3,random_state=123)
## 獲取降維後的資料
tsne_wine_x = tsne.fit_transform(wine_x)
print(tsne_wine_x.shape)
## 在3d空間中視覺化流行降維後的資料空間分布
colors = ["red","blue","green"]
shapes = ["o","s","*"]
fig = plt.figure(figsize=(10,6))
## 將座標系設定為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("特徵1",rotation=20)
ax1.set_ylabel("特徵2",rotation=-20)
ax1.set_zlabel("特徵3",rotation=90)
ax1.azim = 225
ax1.set_title("isomap降維視覺化")
plt.show()
多維尺度分析
## mds進行資料的降維,降維到3維空間中
mds = mds(n_components = 3,dissimilarity = "euclidean",random_state=123)
## 獲取降維後的資料
mds_wine_x = mds.fit_transform(wine_x)
print(mds_wine_x.shape)
## 在3d空間中視覺化流行降維後的資料空間分布
colors = ["red","blue","green"]
shapes = ["o","s","*"]
fig = plt.figure(figsize=(10,6))
## 將座標系設定為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("特徵1",rotation=20)
ax1.set_ylabel("特徵2",rotation=-20)
ax1.set_zlabel("特徵3",rotation=90)
ax1.azim = 225
ax1.set_title("mds降維視覺化")
plt.show()
python用tsne降維 tSNE降維
我有兩套資料訓練和測試。這兩個資料集分別有30213和30235個專案，每個專案有66個維度。在 我正在嘗試應用scikit learn的t sne將維數降到2。由於資料集很大，如果我試圖一次性處理整個資料，我會遇到記憶體錯誤，我會嘗試將它們分成塊，然後一次轉換乙個塊，如下所示 tsne manif...
python 陣列公升維降維
一 numpy.array 增加維度 import numpy as np a np.array 1,2,3 print a 輸出array 1,2,3 print a none 輸出array 1,2,3 print a none 輸出array 1 2 3 二 複製填充並公升維 舉例 把a的sh...
主成分分析 降維
import pandas as pd 引數初始化 inputfile data principal component.xls outputfile tmp dimention reducted.xls 降維後的資料 data pd.read excel inputfile,header none...