Datawhale 集成学习 Task03：掌握偏差与方差理论

先默写本次学习的代码：
一段代码是aic的前向选择，还有两小段是Lasso回归和Ridge回归。
aic指的是赤池信息准则，与其同类还有bic，即贝叶斯信息准则，其属于特征提取的范畴。
前向选择这段代码的作用是，对于一个数据集，有一些特征与自变量相关性不大，过于纠结这些特征会作茧自缚，使得模型的泛化能力弱，不能达到预测的鲁棒性。所以从逐个挑选优质的特征，当特征的增加使得aic的值不是逐渐减小，说明该特征及以后的特征对于预测很可能有反作用，故取其精华去其糟泊。

import numpy as np
import pandas as pd
from sklearn import datasets
from statsmodels.formula.api import ols

boston = datasets.load_boston()
X = boston.data
y = boston.target
features = boston.feature_names
boston_data = pd.DataFrame(X,columns=features)
boston_data['Price'] = y

def forward_select(data,target):
	variate = set(data.columns)
	variate.remove(target)
	selected = []
	current_score,best_new_score = float('inf'),float('inf')
	while variate:
		aic_with_variate = []
		for candidate in variate:
			formula = "{}~{}".format(target,"+".join(selected+[candidate]))
			aic = ols(formula=formula,data=data).fit().aic
			aic_with_variate.append((aic,candidate))
		aic_with_variate.sort(reverse=True)
		best_new_score,best_candidate = aic_with_variate.pop()
		if current_score > best_new_score:
			variate.remove(best_candidate)
			selected.append(best_candidate)
			current_score = best_new_score
			print("aic is {},continuing!".format(current_score))
		else:
			print("selection is over!")
			break
	formula = "{}~{}".format(target,"+".join(selected))
	print("the final formula is {}".format(formula))
	model = ols(formula=formula,data=data).fit()
	return(model)

boston_forward_select_model = forward_select(data=boston_data,target='Price')
boston_forward_select_model.summary()

from sklearn import linear_model
reg_rid = linear_model.Ridge(alpha=0.5)
reg_rid.fit(X,y)
reg_rid.score(X,y)

from sklearn import linear_model
reg_lasso = linear_model.Lasso(alpha=0.5)
reg_lasso.fit(X,y)
reg_lasso.score(X,y)

绘制两幅图，一幅是方差与偏差，另外一幅是Ridge回归和Lasso回归