主成分分析也称主分量分析,旨在利用降维的思想,把多指标转化为少数几个综合指标(即主成分),其中每个主成分都能够反映原始变量的大部分信息,且所含信息互不重复。这种方法在引进多方面变量的同时将复杂因素归结为几个主成分,使问题简单化,同时得到的结果更加科学有效的数据信息。
使用梯度上升法求解主成分
//准备数据
import numpy as np
import matplotlib.pyplot as plt
X = np.empty((100,2))
X[:,0] = np.random.uniform(0.,100.,size=100)
X[:,1] = 0.75*X[:,0]+3.+np.random.normal(0,10.,size=100)
plt.scatter(X[:,0],X[:,1])
plt.show
// 数据demean过程
def demean(X):
return X - np.mean(X,axis=0)
#效用函数
def f(w,X):
return np.sum((X.dot(w)**2))/len(X)
#效用函数导函数
def df_math(w,X):
return X.T.dot(X.dot(w))*2./len(X)
#测试导函数是否正确
def df_debug(w,x,epsilon=0.0001):
res = np.empty(len(w))
for i in range(len(w)):
w_1 = w.copy()
w_1[i] += epsilon
w_2 = w.copy()
w_2[i] -= epsilon
res[i] = (f(w_1,X)-f(w_2,X))/(2*epsilon)
return res
#使w变为单位向量
def direction(w):
return w/np.linalg.norm(w)
#梯度上升
def gradient_ascent(df,X,initial_w,eta,n_iters=1e4,epsilon=1e-8):
cur_iter = 0
w = direction(initial_w)
while cur_iter<n_iters:
gradient = df(w,X)
last_w = w
w = w + eta*gradient
w = direction(w) #每次计算后都应该将w转变为单位向量
if(abs(f(w,X) - f(last_w,X))<epsilon):
break
cur_iter +=1
return w
initial_w = np.random.random(X.shape[1]) #不能使用0向量作为初始向量,因为0向量本身是一个极值点
eta = 0.01
gradient_ascent(df_debug,X,initial_w,eta)
gradient_ascent(df_math,X,initial_w,eta)
求解数据的前n个主成分
def first_n_components(n,X,eta=0.01,n_iters=1e4,epsilon=1e-8):
X_pca = X.copy()
X_pca = demean(X_pca)
res = []
for i in range(n):
initial_w = np.random.random(X_pca.shape[1])
w = first_component(df_math,X_pca,initial_w,eta)
res.append(w)
/*
for i in range(len(X)):
X2[i] = X[i] -X[i].dot(w)*w
*/
X_pca = X_pca - X_pca.dot(w).reshape(-1,1)*w
return res
封装PCA
# _*_ encoding:utf-8 _*_
import numpy as np
class PCA:
def __init__(self,n_components):
self.n_components = n_components
self.components_ = None
def fit(self,X,eta=0.01,n_iters=1e4):
def demean(X):
return X - np.mean(X, axis=0)
# 效用函数
def f(w, X):
return np.sum((X.dot(w) ** 2)) / len(X)
# 效用函数导函数
def df(w, X):
return X.T.dot(X.dot(w)) * 2. / len(X)
def direction(w):
return w / np.linalg.norm(w)
def first_component(df, X, initial_w, eta, n_iters=1e4, epsilon=1e-8):
cur_iter = 0
w = direction(initial_w)
while cur_iter < n_iters:
gradient = df(w, X)
last_w = w
w = w + eta * gradient
w = direction(w) # 每次计算后都应该将w转变为单位向量
if (abs(f(w, X) - f(last_w, X)) < epsilon):
break
cur_iter += 1
return w
def first_n_components(n, X, eta=0.01, n_iters=1e4, epsilon=1e-8):
X_pca = X.copy()
X_pca = demean(X_pca)
res = []
for i in range(n):
initial_w = np.random.random(X_pca.shape[1])
w = first_component(df, X_pca, initial_w, eta)
res.append(w)
X_pca = X_pca - X_pca.dot(w).reshape(-1, 1) * w
return res
X_pca = demean(X)
self.components_ = np.empty(shape=(self.n_components,X.shape[1]))
self.components_ = first_n_components(self.n_components,X)
self.components_ = np.array(self.components_)
return self
def transform(self,X):
return X.dot(self.components_.T)
def inverse_transform(self,X):
return X.dot(self.components_)
def __repr__(self):
return "PCA(n_components=%d)" %self.n_components
scikit-learn中的PCA
scikit-learn中的PCA没有使用梯度上升法求解主成分,因此使用sklearn中的PCA求解的主成分是与我们求解的向量方向是相反的
from sklearn.decomposition import PCA
pca = PCA(n_components=1)
pca.fit(X)
X_transform = pca.transform(X)
X_restore = pca.inverse_transform(X_transform)
plt.scatter(X[:,0],X[:,1],color='b',alpha=0.5)
plt.scatter(X_restore[:,0],X_restore[:,1],color='r',alpha=0.5)
使用PCA处理digits数据集
从图中可以看到,使用PCA将digits数据集的数据维度降低到二维后,knn算法的fit时间降低很多,而score准确率却下降到0.6
pca = PCA(n_components=X_train.shape[1])
pca.fit(X_train)
pca.explained_variance_ratio_ #合并某一维度之后的对数据方差的损失后的正确率
plt.plot([i for i in range(X.shape[1])],[np.sum(pca.explained_variance_ratio_[:i+1]) for i in range(X_train.shape[1])])
pca = PCA(0.95)
pca.fit(X_train)
pca.n_components_ -> pca.n_components = 28
#即降低到28个维度后有原数据95%的正确率