Kaggle: House Prices: Advanced Regression Techniques

Kaggle: House Prices: Advanced Regression Techniques

notebook来自https://www.kaggle.com/neviadomski/how-to-get-to-top-25-with-simple-model-sklearn

思路流程:

1.导入数据,查看数据结构和缺失值情况
重点在于查看缺失值情况的写法:
NAs = pd.concat([train.isnull().sum(), test.isnull().sum()], axis = 1, keys = ['train', 'test']) NAs[NAs.sum(axis=1) > 0]

2.数据预处理(删除无用特征,特征转化,缺失值填充,构造新特征,特征值标准化,转化为dummy)
Q:什么样的特征需要做转化?
A:如某些整型数据只表示类别,其数值本身没有意义,则应转化为dummy
重点学习手动将特征转化为dummy的方法(这里情况稍微还要复杂一点,因为存在同一特征对应两列的情况,如Condition1,Condition2)

3.随机打乱数据,分离训练集和测试集

4.构建多个单一模型

5.模型融合

问题:

1.如何判断一个特征是否是无用特征?

2.模型融合的方法?这里为什是np.exp(GB_model.predict(test_features)) + np.exp(ENS_model.predict(test_features_std))?

3.为什么label分布偏斜需要做转化?

 
In [33]:
#Kaggle: House Prices: Advanced Regression Techniques
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn import ensemble, linear_model, tree
from sklearn.model_selection import train_test_split, cross_val_score
from sklearn.metrics import mean_squared_error, r2_score
from sklearn.utils import shuffle
​
%matplotlib inline
import warnings
warnings.filterwarnings('ignore')
​
train = pd.read_csv('downloads/train.csv')
test = pd.read_csv('downloads/test.csv')
In [8]:
train.head()
Out[8]:
  Id MSSubClass MSZoning LotFrontage LotArea Street Alley LotShape LandContour Utilities ... PoolArea PoolQC Fence MiscFeature MiscVal MoSold YrSold SaleType SaleCondition SalePrice
0 1 60 RL 65.0 8450 Pave NaN Reg Lvl AllPub ... 0 NaN NaN NaN 0 2 2008 WD Normal 208500
1 2 20 RL 80.0 9600 Pave NaN Reg Lvl AllPub ... 0 NaN NaN NaN 0 5 2007 WD Normal 181500
2 3 60 RL 68.0 11250 Pave NaN IR1 Lvl AllPub ... 0 NaN NaN NaN 0 9 2008 WD Normal 223500
3 4 70 RL 60.0 9550 Pave NaN IR1 Lvl AllPub ... 0 NaN NaN NaN 0 2 2006 WD Abnorml 140000
4 5 60 RL 84.0 14260 Pave NaN IR1 Lvl AllPub ... 0 NaN NaN NaN 0 12 2008 WD Normal 250000

5 rows × 81 columns 

In [9]:
#检查缺失值
NAs = pd.concat([train.isnull().sum(), test.isnull().sum()], axis = 1, keys = ['train', 'test']) #sum()默认的axis=0,即跨行
NAs[NAs.sum(axis=1) > 0] #只显示有缺失值的特征
Out[9]:
  train test
Alley 1369 1352.0
BsmtCond 37 45.0
BsmtExposure 38 44.0
BsmtFinSF1 0 1.0
BsmtFinSF2 0 1.0
BsmtFinType1 37 42.0
BsmtFinType2 38 42.0
BsmtFullBath 0 2.0
BsmtHalfBath 0 2.0
BsmtQual 37 44.0
BsmtUnfSF 0 1.0
Electrical 1 0.0
Exterior1st 0 1.0
Exterior2nd 0 1.0
Fence 1179 1169.0
FireplaceQu 690 730.0
Functional 0 2.0
GarageArea 0 1.0
GarageCars 0 1.0
GarageCond 81 78.0
GarageFinish 81 78.0
GarageQual 81 78.0
GarageType 81 76.0
GarageYrBlt 81 78.0
KitchenQual 0 1.0
LotFrontage 259 227.0
MSZoning 0 4.0
MasVnrArea 8 15.0
MasVnrType 8 16.0
MiscFeature 1406 1408.0
PoolQC 1453 1456.0
SaleType 0 1.0
TotalBsmtSF 0 1.0
Utilities 0 2.0
 
In [10]:
#打印R2和RMSE得分
def print_score (prediction, labels):
    print('R2: {}'.format(r2_score(prediction, labels)))
    print('RMSE: {}'.format(np.sqrt(mean_squared_error(prediction, labels))))
​
#对给定的模型进行评估,分别打印训练集上的得分和测试集上的得分
def train_test_score(estimator, x_train, x_test, y_train, y_test):
    train_predictions = estimator.predict(x_train)
    print('------------train-----------')
    print_score(train_predictions, y_train)
    print('------------test------------')
    test_predictions = estimator.predict(x_test)
    print_score(test_predictions, y_test)
In [11]:
#将标签从训练集中分离出来
train_label = train.pop('SalePrice')
​
#将训练集特征和测试集特征拼在一起,便于一起删除无用的特征
features = pd.concat([train, test],  keys = ['train', 'test'])
​
#删除无用特征(为什么说它们是无用特征并没有解释)
features.drop(['Utilities', 'RoofMatl', 'MasVnrArea', 'BsmtFinSF1', 'BsmtFinSF2', 'BsmtUnfSF', 'Heating', 'LowQualFinSF',
               'BsmtFullBath', 'BsmtHalfBath', 'Functional', 'GarageYrBlt', 'GarageArea', 'GarageCond', 'WoodDeckSF',
               'OpenPorchSF', 'EnclosedPorch', '3SsnPorch', 'ScreenPorch', 'PoolArea', 'PoolQC', 'Fence', 'MiscFeature', 'MiscVal'],
              axis=1, inplace=True)
print(features.shape)
(2919, 56)
In [12]:
#将series数据转化为str
#问题:什么样的数据需要转化为str
#答:将原来的某些整型数据转化为str,这些整型数据数字大小本身并没有含义,而只是代表一个类,所以转化为str后,后续再转化为dummy
features['MSSubClass'] = features['MSSubClass'].astype(str)
#pandas调用特征的两种方法:.feature和['feature'],两者效果相同,下面就是.feature方法
features.OverallCond = features.OverallCond.astype(str)
features['KitchenAbvGr'] = features['KitchenAbvGr'].astype(str)
features['YrSold'] = features['YrSold'].astype(str)
features['MoSold'] = features['MoSold'].astype(str)
​
#用众数填充缺失值
features['MSZoning'] = features['MSZoning'].fillna(features['MSZoning'].mode()[0])
features['MasVnrType'] = features['MasVnrType'].fillna(features['MasVnrType'].mode()[0])
features['Electrical'] = features['Electrical'].fillna(features['Electrical'].mode()[0])
features['KitchenQual'] = features['KitchenQual'].fillna(features['KitchenQual'].mode()[0])
features['SaleType'] = features['SaleType'].fillna(features['SaleType'].mode()[0])
​
#用某个特定值填充缺失值
features['LotFrontage'] = features['LotFrontage'].fillna(features['LotFrontage'].mean())
features['Alley'] = features['Alley'].fillna('NOACCESS')
for col in ('BsmtQual', 'BsmtCond', 'BsmtExposure', 'BsmtFinType1', 'BsmtFinType2'):
    features[col] = features[col].fillna('NoBSMT')
features['TotalBsmtSF'] = features['TotalBsmtSF'].fillna(0)
features['FireplaceQu'] = features['FireplaceQu'].fillna('NoFP')
for col in ('GarageType', 'GarageFinish', 'GarageQual'):
    features[col] = features[col].fillna('NoGRG')
features['GarageCars'] = features['GarageCars'].fillna(0.0)
​
#构造新特征
features['TotalSF'] = features['TotalBsmtSF'] + features['1stFlrSF'] + features['2ndFlrSF']
features.drop(['TotalBsmtSF', '1stFlrSF', '2ndFlrSF'], axis=1, inplace=True)
print(features.shape)
(2919, 54)
 
In [13]:
#查看房价分布情况
ax = sns.distplot(train_label)
 
In [14]:
#发现图像整体向左倾斜,所以做log转变
train_label = np.log(train_label)
ax = sns.distplot(train_label)
 
In [15]:
#对数字特征做标准化处理
num_features = features.loc[:,['LotFrontage', 'LotArea', 'GrLivArea', 'TotalSF']]
num_features_standarized = (num_features - num_features.mean()) / num_features.std()
num_features_standarized.head()
Out[15]:
    LotFrontage LotArea GrLivArea TotalSF
train 0 -0.202033 -0.217841 0.413476 0.022999
1 0.501785 -0.072032 -0.471810 -0.029167
2 -0.061269 0.137173 0.563659 0.196886
3 -0.436639 -0.078371 0.427309 -0.092511
4 0.689469 0.518814 1.377806 0.988072
 
In [16]:
ax = sns.pairplot(num_features_standarized)
 
In [17]:
#重点
#convert categorical data to dummies
#将所有condition不重复的记录在一个set中
conditions = set([x for x in features['Condition1']] + [x for x in features['Condition2']]) 
#自定义dummy变量,行数为阳历数,列数为原condition数据转化为dummy后的维数
dummies = pd.DataFrame(data = np.zeros((len(features.index), len(conditions))), index = features.index, columns = conditions)
#遍历所有样例,将原来的condition信息转化为对应的dummy信息
for i, cond in enumerate(zip(features['Condition1'], features['Condition2'])):
#用ix找到位置,注意cond可能包含Condition1和Condition2两个位置的信息,对应dummies数组的两个点,所以需要用ix而不能简单的直接用dummies[i,cond]
    dummies.ix[i, cond] = 1
#将dummy后的特征数据拼接到原features后面,并给dummy特征的index增加前缀
features = pd.concat([features, dummies.add_prefix('Cond_')], axis = 1)
#最后就可以删除原来的Condition特征
features.drop(['Condition1', 'Condition2'], axis = 1, inplace =True)
print(features.shape)
(2919, 61)
 
In [18]:
features.head()
 
Out[18]:
    Id MSSubClass MSZoning LotFrontage LotArea Street Alley LotShape LandContour LotConfig ... TotalSF Cond_PosA Cond_Artery Cond_PosN Cond_RRAn Cond_RRAe Cond_Feedr Cond_Norm Cond_RRNn Cond_RRNe
train 0 1 60 RL 65.0 8450 Pave NOACCESS Reg Lvl Inside ... 2566.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0
1 2 20 RL 80.0 9600 Pave NOACCESS Reg Lvl FR2 ... 2524.0 0.0 0.0 0.0 0.0 0.0 1.0 1.0 0.0 0.0
2 3 60 RL 68.0 11250 Pave NOACCESS IR1 Lvl Inside ... 2706.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0
3 4 70 RL 60.0 9550 Pave NOACCESS IR1 Lvl Corner ... 2473.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0
4 5 60 RL 84.0 14260 Pave NOACCESS IR1 Lvl FR2 ... 3343.0 0.0 0.0 0.0 0.0 0.0 0.0 1.0 0.0 0.0

5 rows × 61 columns

 
In [19]:
#convert Exterior to dummies
Exterior = set([x for x in features['Exterior1st']] + [x for x in features['Exterior2nd']])
dummies = pd.DataFrame(data = np.zeros([len(features.index), len(Exterior)]), index = features.index, columns = Exterior)
for i, ext in enumerate(zip(features['Exterior1st'], features['Exterior2nd'])):
    dummies.ix[i, ext] = 1
features = pd.concat([features, dummies.add_prefix('Ext_')], axis = 1)
features.drop(['Exterior1st', 'Exterior2nd', 'Ext_nan'], axis = 1, inplace = True)
print(features.shape)
(2919, 78)
 
In [20]:
features.dtypes[features.dtypes == 'object'].index
 
Out[20]:
Index(['MSSubClass', 'MSZoning', 'Street', 'Alley', 'LotShape', 'LandContour',
       'LotConfig', 'LandSlope', 'Neighborhood', 'BldgType', 'HouseStyle',
       'OverallCond', 'RoofStyle', 'MasVnrType', 'ExterQual', 'ExterCond',
       'Foundation', 'BsmtQual', 'BsmtCond', 'BsmtExposure', 'BsmtFinType1',
       'BsmtFinType2', 'HeatingQC', 'CentralAir', 'Electrical', 'KitchenAbvGr',
       'KitchenQual', 'FireplaceQu', 'GarageType', 'GarageFinish',
       'GarageQual', 'PavedDrive', 'MoSold', 'YrSold', 'SaleType',
       'SaleCondition'],
      dtype='object')
 
In [21]:
#遍历特定类型数据的方法:for col in features.dtypes[features.dtypes == 'object'].index
#convert all other categorical vars to dummies
for col in features.dtypes[features.dtypes == 'object'].index:
    for_dummy = features.pop(col)
    features = pd.concat([features, pd.get_dummies(for_dummy, prefix = col)], axis = 1)
print(features.shape)
(2919, 263)
 
In [22]:
#用之前几个标准化的数据更新features
​
features_standardized = features.copy()
features_standardized.update(num_features_standarized)
In [23]:
#重新分离训练集和测试集
#首先分离没有标准化的features
train_features = features.loc['train'].drop(['Id'], axis=1).select_dtypes(include=[np.number]).values
test_features = features.loc['test'].drop(['Id'], axis=1).select_dtypes(include=[np.number]).values
​
#再分离标准化的数据
train_features_std = features_standardized.loc['train'].drop(['Id'], axis=1).select_dtypes(include=[np.number]).values
test_features_std = features_standardized.loc['test'].drop(['Id'], axis=1).select_dtypes(include=[np.number]).values
print(train_features.shape)
print(train_features_std.shape)
(1460, 262)
(1460, 262)
 
In [24]:
#shuffle train dataset
train_features_std, train_features, train_label = shuffle(train_features_std, train_features, train_label, random_state = 5)
In [25]:
#split train and test data
x_train, x_test, y_train, y_test = train_test_split(train_features, train_label, test_size = 0.1, random_state = 200)
x_train_std, x_test_std, y_train_std, y_test_std = train_test_split(train_features_std, train_label, test_size = 0.1, random_state = 200)
In [26]:
#构建第一个模型:ElasticNet
ENSTest = linear_model.ElasticNetCV(alphas=[0.0001, 0.0005, 0.001, 0.01, 0.1, 1, 10], l1_ratio=[.01, .1, .5, .9, .99], max_iter=5000).fit(x_train_std, y_train_std)
train_test_score(ENSTest, x_train_std, x_test_std, y_train_std, y_test_std)
------------train-----------
R2: 0.9009283127352861
RMSE: 0.11921419084690392
------------test------------
R2: 0.8967299522701895
RMSE: 0.11097042840114624
 
In [27]:
#测试模型的交叉验证得分
score = cross_val_score(ENSTest, train_features_std, train_label, cv = 5)
print('Accurary: %0.2f +/- %0.2f' % (score.mean(), score.std()*2))
Accurary: 0.88 +/- 0.10
 
In [28]:
#构建第二个模型:GradientBoosting
GB = ensemble.GradientBoostingRegressor(n_estimators=3000, learning_rate = 0.05, max_depth = 3, max_features = 'sqrt', 
                                        min_samples_leaf = 15,
                                        min_samples_split = 10, loss = 'huber').fit(x_train_std, y_train_std)
train_test_score(GB, x_train_std, x_test_std, y_train_std, y_test_std)
------------train-----------
R2: 0.9607778449577035
RMSE: 0.07698826081848897
------------test------------
R2: 0.9002871760789876
RMSE: 0.10793269100940146
 
In [29]:
#构建第二个模型:GradientBoosting
GB = ensemble.GradientBoostingRegressor(n_estimators=3000, learning_rate = 0.05, max_depth = 3, max_features = 'sqrt', 
                                        min_samples_leaf = 15,
                                        min_samples_split = 10, loss = 'huber').fit(x_train_std, y_train_std)
train_test_score(GB, x_train_std, x_test_std, y_train_std, y_test_std)
Accurary: 0.90 +/- 0.04
 
In [30]:
#模型融合
GB_model = GB.fit(train_features, train_label)
ENS_model = ENSTest.fit(train_features_std, train_label)
In [31]:
#为什么模型融合公式是这样的?
Final_score = (np.exp(GB_model.predict(test_features)) + np.exp(ENS_model.predict(test_features_std))) / 2
 
In [32]:
#写入csv文件
pd.DataFrame({'Id':test.Id, 'SalePrice':Final_score}).to_csv('submit.csv', index=False)

转载于:https://www.cnblogs.com/RB26DETT/p/11566650.html

你可能感兴趣的:(人工智能,数据结构与算法,python)