用Python做一个房价预测小工具！

train = pd.read_csv('./data/train.csv')
test = pd.read_csv('./data/test.csv')

sns.distplot(train['SalePrice']);

# 计算列之间相关性
corrmat = train.corr()
# 取 top10
k = 10
cols = corrmat.nlargest(k, 'SalePrice')['SalePrice'].index
# 绘图
cm = np.corrcoef(train[cols].values.T)
sns.set(font_scale=1.25)
hm = sns.heatmap(cm, cbar=True, annot=True, square=True, fmt='.2f', annot_kws={'size': 10}, yticklabels=cols.values, xticklabels=cols.values)
plt.show()

# 获取数值型特征
numeric_features = train.dtypes[train.dtypes != 'object'].index
# 计算每个特征的离群样本
for feature in numeric_features:
outs = detect_outliers(train[feature], train['SalePrice'],top=5, plot=False)
all_outliers.extend(outs)
# 输出离群次数最多的样本
print(Counter(all_outliers).most_common())
# 剔除离群样本
train = train.drop(train.index[outliers])

y = train.SalePrice.reset_index(drop=True)
train_features = train.drop(['SalePrice'], axis=1)
test_features = test
features = pd.concat([train_features, test_features]).reset_index(drop=True)

features['MSSubClass'] = features['MSSubClass'].apply(str)
features['YrSold'] = features['YrSold'].astype(str)
features['MoSold'] = features['MoSold'].astype(str)

# Functional：文档提供了典型值 Typ
features['Functional'] = features['Functional'].fillna('Typ') #Typ 是典型值
# 分组填充需要按照相似的特征分组，取众数或中位数
# MSZoning（房屋区域）按照 MSSubClass（房屋）类型分组填充众数
features['MSZoning'] = features.groupby('MSSubClass')['MSZoning'].transfORM(lambda x: x.fillna(x.mode()[0]))
#LotFrontage（到接到举例）按Neighborhood分组填充中位数
features['LotFrontage'] = features.groupby('Neighborhood')['LotFrontage'].transform(lambda x: x.fillna(x.median()))
# 车库相关的数值型特征，空代表无，使用0填充空值。
for col in ('GarageYrBlt', 'GarageArea', 'GarageCars'):
features[col] = features[col].fillna(0)

# skew()方法，计算特征的偏度（skewness）。
skew_features = features[numeric_features].apply(lambda x: skew(x)).sort_values(ascending=False)
# 取偏度大于 0.15 的特征
high_skew = skew_features[skew_features > 0.15]
skew_index = high_skew.index
# 处理高偏度特征，将其转化为正态分布，也可以使用简单的log变换
for i in skew_index:
features[i] = boxcox1p(features[i], boxcox_normmax(features[i] + 1))

features = features.drop(['Utilities', 'Street', 'PoolQC',], axis=1)

# 将原施工日期和改造日期融合
features['YrBltAndRemod']=features['YearBuilt']+features['YearRemodAdd']
# 将地下室面积、1楼、2楼面积融合
features['TotalSF']=features['TotalBsmtSF'] + features['1stFlrSF'] + features['2ndFlrSF']

features['haspool'] = features['PoolArea'].apply(lambda x: 1 if x > 0 else 0)
features['has2ndfloor'] = features['2ndFlrSF'].apply(lambda x: 1 if x > 0 else 0)

X = features.iloc[:len(y), :]
X_sub = features.iloc[len(y):, :]
X = np.array(X.copy())
y = np.array(y)
X_sub = np.array(X_sub.copy())

from sklearn.linear_model import RidgeCV
from sklearn.pipeline import make_pipeline
from sklearn.model_selection import KFold
kfolds = KFold(n_splits=10, shuffle=True, random_state=42)
alphas_alt = [14.5, 14.6, 14.7, 14.8, 14.9, 15, 15.1, 15.2, 15.3, 15.4, 15.5]
ridge = make_pipeline(RobustScaler(), RidgeCV(alphas=alphas_alt, cv=kfolds))

# lasso
lasso = make_pipeline(
RobustScaler(),
LassoCV(max_iter=1e7, alphas=alphas2, random_state=42, cv=kfolds))
#elastic net
elasticnet = make_pipeline(
RobustScaler(),
ElasticNetCV(max_iter=1e7, alphas=e_alphas, cv=kfolds, l1_ratio=e_l1ratio))
#svm
svr = make_pipeline(RobustScaler(), SVR(
C=20,
epsilon=0.008,
gamma=0.0003,
))
#GradientBoosting（展开到一阶导数）
gbr = GradientBoostingRegressor(...)
#lightgbm
lightgbm = LGBMRegressor(...)
#xgboost（展开到二阶导数）
xgboost = XGBRegressor(...)

#模型评分函数
def cv_rmse(model, X=X):
rmse = np.sqrt(-cross_val_score(model, X, y, scoring="neg_mean_squared_error", cv=kfolds))
return (rmse)

score = cv_rmse(ridge)
print("Ridge score: {:.4f} ({:.4f})n".format(score.mean(), score.std()), datetime.now(), ) #0.1024

# 训练模型
ridge.fit(X, y)
# 模型预测
submission.iloc[:,1] = np.floor(np.expm1(ridge.predict(X_sub)))
# 输出测试结果
submission = pd.read_csv("./data/sample_submission.csv")
submission.to_csv("submission_single.csv", index=False)

stack_gen =
StackingCVRegressor(
regressors=(ridge, lasso, elasticnet, gbr, xgboost, lightgbm),
meta_regressor=xgboost,
use_features_in_secondary=True)

# 训练单个模型
ridge_model_full_data = ridge.fit(X, y)
lasso_model_full_data = lasso.fit(X, y)
elastic_model_full_data = elasticnet.fit(X, y)
gbr_model_full_data = gbr.fit(X, y)
xgb_model_full_data = xgboost.fit(X, y)
lgb_model_full_data = lightgbm.fit(X, y)
svr_model_full_data = svr.fit(X, y)
models = [
ridge_model_full_data, lasso_model_full_data, elastic_model_full_data,
gbr_model_full_data, xgb_model_full_data, lgb_model_full_data,
svr_model_full_data, stack_gen_model
]
# 分配模型权重
public_coefs = [0.1, 0.1, 0.1, 0.1, 0.15, 0.1, 0.1, 0.25]
# 线性融合，取加权平均
def linear_blend_models_predict(data_x,models,coefs, bias):
tmp=[model.predict(data_x) for model in models]
tmp = [c*d for c,d in zip(coefs,tmp)]
pres=np.array(tmp).swapaxes(0,1)
pres=np.sum(pres,axis=1)
return pres