机器学习入门线性回归 岭回归与Lasso回归(二)
一 线性回归(Linear Regression )
1. 线性回归概述
回归的目的是预测数值型数据的目标值,最直接的方法就是根据输入写出一个求出目标值的计算公式,也就是所谓的回归方程,例如y = ax1+bx2,其中求回归系数的过程就是回归。那么回归是如何预测的呢?当有了这些回归系数,给定输入,具体的做法就是将回归系数与输入相乘,再将结果加起来就是最终的预测值。说到回归,一般指的都是线性回归,当然也存在非线性回归,在此不做讨论。
假定输入数据存在矩阵x中,而回归系数存放在向量w中。那么对于给定的数据x1,预测结果可以通过y1 = x1Tw给出,那么问题就是来寻找回归系数。一个最常用的方法就是寻找误差最小的w,误差可以用预测的y值和真实的y值的差值表示,由于正负差值的差异,可以选用平方误差,也就是对预测的y值和真实的y值的平方求和,用矩阵可表示为:
(y - xw)T(y - xw)
\]
现在问题就转换为寻找使得上述矩阵值最小的w,对w求导为:xT(y - xw),令其为0,解得:
w = (xTx)-1xTy
\]
这就是采用此方法估计出来的.
案例: 糖尿病回归分析
import numpy as np
import pandas as pd
from pandas import Series,DataFrame
import matplotlib.pyplot as plt
%matplotlib inline
#导包读取
from sklearn import datasets
diabetes = datasets.load_diabetes()
#生成DataFrame 与 Series对象
train = DataFrame(data = diabetes.data,columns = diabetes.feature_names)
target = diabetes.target
# 数据集拆分 训练集和样本集
# 模型选择的包
from sklearn.model_selection import train_test_split
## train 数据样本集
# target 样本标签
# test_size 测试集的比例
# random_state 随机数种子,限定随机取值的随机顺序,每一个种子固定一组随机数
X_train, X_test, y_train, y_test = train_test_split(train, target, test_size=0.1, random_state=42)
#导入线性回归与KNN模型
from sklearn.linear_model import LinearRegression
from sklearn.neighbors import KNeighborsRegressor
#训练数据
linear = LinearRegression()
linear.fit(X_train,y_train)
#预测数据
y_ = linear.predict(X_test)
plt.plot(y_,label='Predict')
plt.plot(y_test,label='True')
plt.legend()
aaarticlea/png;base64,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" alt="" />
# 线性回归模型的拟合度的好坏,就是看真实值和预测值之间的误差的大小
# 残差直方图,评价回归模型的好坏,瘦高就是好,矮胖就是不好
plt.hist(y_test - y_,rwidth=0.9,bins=10)
plt.xlabel('cost-value')
plt.ylabel('numbers')
aaarticlea/png;base64,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" alt="" />
# 还可以用r2_score来评估回归模型的好坏,拟合优度
# 比较不同的回归模型的r2的分值大小,越大越好
from sklearn.metrics import r2_score
r2_score(y_test,y_)
#0.5514251914993504
# 使用KNN回归器处理
knn = KNeighborsRegressor(n_neighbors=7)
knn.fit(X_train,y_train)
knn_y_ = knn.predict(X_test)
plt.plot(y_,label='Linear-Predict')
plt.plot(knn_y_,label='knn-Predict')
plt.plot(y_test,label='True')
plt.legend()
aaarticlea/png;base64,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" alt="" />
# 先用残差直方图比较KNN和Linear哪个更好
plt.figure(figsize=(10,4))
axes1 = plt.subplot(1,2,1)
axes1.hist(knn_y_ - y_test,rwidth=0.9)
axes1.set_xlabel('cost-value')
axes1.set_ylabel('numbers')
axes1.set_title('KNN')
axes2 = plt.subplot(1,2,2)
axes2.hist(y_test - y_,rwidth=0.9,bins=10)
axes2.set_xlabel('cost-value')
axes2.set_ylabel('numbers')
axes2.set_title('Linear')
aaarticlea/png;base64,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" alt="" />
注意:
np.random.seed(1)
np.random.randint(0,10,size=10)array([5, 8, 9, 5, 0, 0, 1, 7, 6, 9])
二 局部加权线性回归(Locally Weighted Linear Regression,LWLR)
1.概述
针对于线性回归存在的欠拟合现象,可以引入一些偏差得到局部加权线性回归对算法进行优化。在该算法中,给待测点附近的每个点赋予一定的权重,进而在所建立的子集上进行给予最小均方差来进行普通的回归,分析可得回归系数w可表示为:
w = (xTWx)-1xTWy,
\]
其中W为每个数据点赋予的权重,那么怎样求权重呢,核函数可以看成是求解点与点之间的相似度,在此可以采用核函数,相应的根据预测点与附近点之间的相似程度赋予一定的权重,在此选用最常用的高斯核,则权重可以表示为:
w(i,i) = exp(|x(i) - x| / -2k2),
\]
其中K为宽度参数,至于此参数的取值,目前仍没有一个确切的标准,只有一个范围的描述,所以在算法的应用中,可以采用不同的取值分别调试,进而选取最好的结果。
三 岭回归
1.概述
为了解决上述问题,统计学家引入了“岭回归”的概念。简单说来,岭回归就是在矩阵XTX上加上一个λr,从而使得矩阵非奇异,从而能对XTX + λx求逆。其中矩阵r为一个m*m的单位矩阵,对角线上的元素全为1,其他元素全为0,而λ是一个用户定义的数值,这种情况下,回归系数的计算公式将变为:
w = (xTx+λI)-1xTy,
\]
其中I是一个单位矩阵。
岭回归就是用了单位矩阵乘以常量λ,因为只I贯穿了整个对角线,其余元素为0,形象的就是在0构成的平面上有一条1组成的“岭”,这就是岭回归中岭的由来。
岭回归最先是用来处理特征数多与样本数的情况,现在也用于在估计中加入偏差,从而得到更好的估计。这里引入λ限制了所有w的和,通过引入该惩罚项,能够减少不重要的参数,这个技术在统计学上也叫做缩减。缩减方法可以去掉不重要的参数,因此能更好的理解数据。选取不同的λ进行测试,最后得到一个使得误差最小λ。
**优点
机器学习入门线性回归 岭回归与Lasso回归(二)的更多相关文章
- 【机器学习】正则化的线性回归 —— 岭回归与Lasso回归
注:正则化是用来防止过拟合的方法.在最开始学习机器学习的课程时,只是觉得这个方法就像某种魔法一样非常神奇的改变了模型的参数.但是一直也无法对其基本原理有一个透彻.直观的理解.直到最近再次接触到这个概念 ...
- 机器学习之五 正则化的线性回归-岭回归与Lasso回归
机器学习之五 正则化的线性回归-岭回归与Lasso回归 注:正则化是用来防止过拟合的方法.在最开始学习机器学习的课程时,只是觉得这个方法就像某种魔法一样非常神奇的改变了模型的参数.但是一直也无法对其基 ...
- 多元线性回归模型的特征压缩:岭回归和Lasso回归
多元线性回归模型中,如果所有特征一起上,容易造成过拟合使测试数据误差方差过大:因此减少不必要的特征,简化模型是减小方差的一个重要步骤.除了直接对特征筛选,来也可以进行特征压缩,减少某些不重要的特征系数 ...
- 线性回归大结局(岭(Ridge)、 Lasso回归原理、公式推导),你想要的这里都有
本文已参与「新人创作礼」活动,一起开启掘金创作之路. 线性模型简介 所谓线性模型就是通过数据的线性组合来拟合一个数据,比如对于一个数据 \(X\) \[X = (x_1, x_2, x_3, ..., ...
- 岭回归和lasso回归(转)
回归和分类是机器学习算法所要解决的两个主要问题.分类大家都知道,模型的输出值是离散值,对应着相应的类别,通常的简单分类问题模型输出值是二值的,也就是二分类问题.但是回归就稍微复杂一些,回归模型的输出值 ...
- 岭回归和Lasso回归以及norm1和norm2
norm代表的是距离,两个向量的距离:下图代表的就是p-norm,其实是对向量里面元素的一种运算: 最简单的距离计算(规范)是欧式距离(Euclidean distance),两点间距离是如下来算的, ...
- 岭回归与Lasso回归
线性回归的一般形式 过拟合问题及其解决方法 问题:以下面一张图片展示过拟合问题 解决方法:(1):丢弃一些对我们最终预测结果影响不大的特征,具体哪些特征需要丢弃可以通过PCA算法来实现:(2):使用正 ...
- 回归算法比较(线性回归,Ridge回归,Lasso回归)
代码: # -*- coding: utf-8 -*- """ Created on Mon Jul 16 09:08:09 2018 @author: zhen &qu ...
- 吴裕雄 数据挖掘与分析案例实战(7)——岭回归与LASSO回归模型
# 导入第三方模块import pandas as pdimport numpy as npimport matplotlib.pyplot as pltfrom sklearn import mod ...
随机推荐
- CentOS7编译curl
1.下载curl源代码 https://curl.haxx.se/download.html 2.进入curl目录 ./configure --prefix=/usr/local/curl make ...
- push到github报错解决方法
在push代码到远程仓库时,报了如下的错误: $ git push -u origin master To https://github.com/11pdg/group-buy.git ! [reje ...
- MySQL字符串函数:字符串截取
MySQL 字符串截取函数:left(), right(), substring(), substring_index().还有 mid(), substr().其中,mid(), substr() ...
- 《Centos服务器版安装教程》
安装前准备: (1) 首先大家需要在电脑上安装一个VMware (2) Centos7系列的一个服务器版镜像 有了这两样东西,下面我们就开始安装了 一. 打开VMware,新建一个虚拟机 ...
- D3基础--数轴
转载请注明出处! 概述: 与比例尺类似,D3的数轴实际上也使用来定义参数的函数.但与比例尺不同的是,调用数轴函数并不会返回值,而是会生成数轴相关的可见元素.包括:轴线,标签和刻度. 但是要注意数轴函数 ...
- jQuery事件篇---过滤选择器 & 表单选择器
内容提纲: 过滤选择器 1.基本过滤器 2.内容过滤器 3.可见性过滤器 4.子元素过滤器 5.其他方法 表单选择器 6.常规选择器 7.表单选择器 8.表单过滤器 发文不易,转载请注明出处! 过滤选 ...
- swagger api文档添加jwt授权配置
最近写的swagger文档,要加jwt授权,所以几经google终于搞定了,简简单单几行配置如下: securityDefinitions: APIKey: type: apiKey name: Au ...
- Html.DropDownList传值
Html.DropDownList传值: 可以传入明确的IEnumerable<SelectListItem>,也可以通过ViewBag或者ViewData隐式地传入,前提是需要相同的名称 ...
- 项目中遇到的问题——jsp:include
昨晚记错了,项目中用的是这个<jsp:attribute>,不过没关系,都差不多!原理是传参数 具体用法: 假设有两个tag文件 aaa 和 bbb aaa有两个属性:name age ...
- hdu 3999 二叉查找树
The order of a Tree Time Limit: 2000/1000 MS (Java/Others) Memory Limit: 32768/32768 K (Java/Othe ...