机器学习加利福尼亚房价预测

提示&＃xff1a;导入包

例如&＃xff1a;

import pandas as pd
import numpy as np
from sklearn.datasets import fetch_california_housing
from sklearn.model_selection import KFold, train_test_split
from sklearn.metrics import mean_squared_error as mse
from xgboost import XGBRegressor
from lightgbm import LGBMRegressor
from folium import Map
from folium.plugins import HeatMap
import matplotlib.pyplot as plt
%matplotlib inline
%config InlineBackend.figure_format &＃61; &＃39;retina&＃39;
import seaborn as sns
sns.set_style(&＃39;whitegrid&＃39;)
!pip install -q flaml
import flaml
!pip install -q autogluon
from autogluon.tabular import TabularPredictor


提示&＃xff1a;这里可以添加要学的内容

df_train &＃61; pd.read_csv(&＃39;/kaggle/input/playground-series-s3e1/train.csv&＃39;, index_col&＃61;0)
df_test &＃61; pd.read_csv(&＃39;/kaggle/input/playground-series-s3e1/test.csv&＃39;, index_col&＃61;0)
features, target &＃61; df_train.columns[:-1], df_train.columns[-1]


特征构造

original &＃61; fetch_california_housing()
assert original[&＃39;feature_names&＃39;] &＃61;&＃61; list(features)
assert original[&＃39;target_names&＃39;][0] &＃61;&＃61; target
df_original &＃61; pd.DataFrame(original[&＃39;data&＃39;], columns&＃61;features)
df_original[target] &＃61; original[&＃39;target&＃39;]


提示&＃xff1a;Missing values

df_train.isna().sum()


查看分布

ncols &＃61; 3
nrows &＃61; np.ceil(len(df_train.columns)/ncols).astype(int)
cols &＃61; df_train.corrwith(df_train[target]).abs().sort_values(ascending&＃61;False)
fig, axs &＃61; plt.subplots(ncols&＃61;ncols, nrows&＃61;nrows, figsize&＃61;(10,nrows*2))
for idx, (c, corr) in enumerate(cols.items()):
row &＃61; idx // ncols
col &＃61; idx % ncols
sns.histplot(df_train, x&＃61;c, ax&＃61;axs[row, col])
axs[row,col].set_xlabel(f&＃39;{c}, corr &＃61; {corr:.3f}&＃39;)
plt.tight_layout()


Some observations:

House value is capped
MedInc has the highest correlation with house value
Look at the data on a map
Credit: https://www.kaggle.com/code/jcaliz/ps-s03e01-a-complete-eda

heat_data &＃61; [[row[&＃39;Latitude&＃39;],row[&＃39;Longitude&＃39;]] for _, row in df_train.iterrows()]
heat_map &＃61; Map(df_train[[&＃39;Latitude&＃39;, &＃39;Longitude&＃39;]].mean(axis&＃61;0), zoom_start&＃61;6)
HeatMap(heat_data, radius&＃61;10).add_to(heat_map)
heat_map


提示&＃xff1a;这里统计学习计划的总量

例如&＃xff1a;

df_train[&＃39;source&＃39;] &＃61; &＃39;train&＃39;
df_test[&＃39;source&＃39;] &＃61; &＃39;test&＃39;
df_original[&＃39;source&＃39;] &＃61; &＃39;original&＃39;
df_mixed &＃61; pd.concat([df_train, df_test, df_original])


fig, axs &＃61; plt.subplots(ncols&＃61;ncols, nrows&＃61;nrows, figsize&＃61;(15,nrows*2))
for idx, (c, corr) in enumerate(cols.items()):
row &＃61; idx // ncols
col &＃61; idx % ncols
sns.boxplot(data&＃61;df_mixed, x&＃61;c, y&＃61;&＃39;source&＃39;, ax&＃61;axs[row, col])
plt.tight_layout()


Baseline with gradient boosting

X_train, y_train &＃61; df_train[features].values, df_train[target].values
X_test &＃61; df_test[features].values


oof &＃61; np.zeros(len(df_train))
models &＃61; []
for fold, (idx_tr, idx_vl) in enumerate(cv.split(X_train)):
X_tr, y_tr &＃61; X_train[idx_tr], y_train[idx_tr]
X_vl, y_vl &＃61; X_train[idx_vl], y_train[idx_vl]
model &＃61; LGBMRegressor()
model.fit(X_tr, y_tr)
oof[idx_vl] &＃61; model.predict(X_vl)
models.append(model)

r &＃61; mse(y_vl, oof[idx_vl], squared&＃61;False)
print(f&＃39;Fold {fold} rmse: {r:.4}&＃39;)
print(f&＃39;OOF rmse: {mse(y_train, oof, squared&＃61;False):.4}&＃39;)


Fold 0 rmse: 0.5708
Fold 1 rmse: 0.5685
Fold 2 rmse: 0.5595
Fold 3 rmse: 0.5704
Fold 4 rmse: 0.5764
OOF rmse: 0.5692

Adding original data

df_combined &＃61; pd.concat([df_train, df_original]).drop(columns&＃61;[&＃39;source&＃39;])
X_train, y_train &＃61; df_combined[features].values, df_combined[target].values
oof &＃61; np.zeros(len(X_train))
models &＃61; []
for fold, (idx_tr, idx_vl) in enumerate(cv.split(X_train)):
X_tr, y_tr &＃61; X_train[idx_tr], y_train[idx_tr]
X_vl, y_vl &＃61; X_train[idx_vl], y_train[idx_vl]
model &＃61; LGBMRegressor()
model.fit(X_tr, y_tr)
oof[idx_vl] &＃61; model.predict(X_vl)
models.append(model)

r &＃61; mse(y_vl, oof[idx_vl], squared&＃61;False)
print(f&＃39;Fold {fold} rmse: {r:.4}&＃39;)
print(f&＃39;OOF rmse: {mse(y_train, oof, squared&＃61;False):.4}&＃39;)


Fold 0 rmse: 0.53
Fold 1 rmse: 0.5367
Fold 2 rmse: 0.5385
Fold 3 rmse: 0.5372
Fold 4 rmse: 0.5411
OOF rmse: 0.5367

参考文章&＃xff1a;https://www.kaggle.com/code/phongnguyen1/s03e01-original-data-boost-automl