In [1]:
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
In [2]:
# 导入数据
df_train = pd.read_csv("hour.csv")
df_train.head()
Out[2]:
| instant | dteday | season | yr | mnth | hr | holiday | weekday | workingday | weathersit | temp | atemp | hum | windspeed | casual | registered | cnt | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | 2011-01-01 | 1 | 0 | 1 | 0 | 0 | 6 | 0 | 1 | 0.24 | 0.2879 | 0.81 | 0.0 | 3 | 13 | 16 |
| 1 | 2 | 2011-01-01 | 1 | 0 | 1 | 1 | 0 | 6 | 0 | 1 | 0.22 | 0.2727 | 0.80 | 0.0 | 8 | 32 | 40 |
| 2 | 3 | 2011-01-01 | 1 | 0 | 1 | 2 | 0 | 6 | 0 | 1 | 0.22 | 0.2727 | 0.80 | 0.0 | 5 | 27 | 32 |
| 3 | 4 | 2011-01-01 | 1 | 0 | 1 | 3 | 0 | 6 | 0 | 1 | 0.24 | 0.2879 | 0.75 | 0.0 | 3 | 10 | 13 |
| 4 | 5 | 2011-01-01 | 1 | 0 | 1 | 4 | 0 | 6 | 0 | 1 | 0.24 | 0.2879 | 0.75 | 0.0 | 0 | 1 | 1 |
In [3]:
df_train.isnull().sum()
Out[3]:
instant 0 dteday 0 season 0 yr 0 mnth 0 hr 0 holiday 0 weekday 0 workingday 0 weathersit 0 temp 0 atemp 0 hum 0 windspeed 0 casual 0 registered 0 cnt 0 dtype: int64
In [4]:
# 数据集更新了，年月日已经拆分了
# df_train["year"] = pd.DatatimeIndex(df_train["datatime"]).year
In [5]:
fig, axs = plt.subplots(2,2, sharey=True)
df_train.groupby("weathersit").plot(y="cnt", marker="o", ax=axs[0,0])
df_train.groupby("hum").mean().plot(y="cnt", marker="o",figsize=(24,16), ax=axs[0,1])
df_train.groupby("temp").mean().plot(y="cnt", marker="o", ax=axs[1,0])
df_train.groupby("windspeed").plot(y="cnt", marker="o", ax=axs[1,1])
plt.show
Out[5]:
<function matplotlib.pyplot.show(*args, **kw)>
In [6]:
# 从上面数据看到有的特征不求均值找不到规律
In [7]:
fig, axs = plt.subplots(2,2, sharey=True)
df_train.groupby("weathersit").mean().plot(y="cnt", marker="o", ax=axs[0,0])
df_train.groupby("hum").mean().plot(y="cnt", marker="o",figsize=(24,16), ax=axs[0,1])
df_train.groupby("temp").mean().plot(y="cnt", marker="o", ax=axs[1,0])
df_train.groupby("windspeed").mean().plot(y="cnt", marker="o", ax=axs[1,1])
plt.show
Out[7]:
<function matplotlib.pyplot.show(*args, **kw)>
In [8]:
df_train.groupby("hr").mean().plot(y="cnt", marker="o")
plt.title("mean cnt per hour")
plt.show()
In [9]:
fig,axs = plt.subplots(3,3, sharey= True)
df_train.plot(x="temp", y="cnt", kind="scatter",figsize=(24,16),ax=axs[0,0],color="magenta")
df_train.plot(x="weekday", y="cnt", kind="scatter",ax=axs[0,1],color="cyan")
df_train.plot(x="hum", y="cnt", kind="scatter",ax=axs[0,2],color="bisque")
df_train.plot(x="windspeed", y="cnt", kind="scatter",ax=axs[1,0],color="coral")
df_train.plot(x="mnth", y="cnt", kind="scatter",ax=axs[1,1],color="darkblue")
df_train.plot(x="hr", y="cnt", kind="scatter",ax=axs[1,2],color="deeppink")
df_train.plot(x="holiday", y="cnt", kind="scatter",ax=axs[2,0],color="cyan")
df_train.plot(x="workingday", y="cnt", kind="scatter",ax=axs[2,1],color="cyan")
column = df_train.columns
corr = df_train[column].corr()
plt.figure()
# 特征间的相关矩阵
plt.matshow(corr)
plt.colorbar()
plt.show()
<Figure size 432x288 with 0 Axes>
In [10]:
df = df_train.drop(["dteday"], axis=1)
In [11]:
df.head(3)
Out[11]:
| instant | season | yr | mnth | hr | holiday | weekday | workingday | weathersit | temp | atemp | hum | windspeed | casual | registered | cnt | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | 1 | 0 | 1 | 0 | 0 | 6 | 0 | 1 | 0.24 | 0.2879 | 0.81 | 0.0 | 3 | 13 | 16 |
| 1 | 2 | 1 | 0 | 1 | 1 | 0 | 6 | 0 | 1 | 0.22 | 0.2727 | 0.80 | 0.0 | 8 | 32 | 40 |
| 2 | 3 | 1 | 0 | 1 | 2 | 0 | 6 | 0 | 1 | 0.22 | 0.2727 | 0.80 | 0.0 | 5 | 27 | 32 |
In [12]:
name = df.drop(['cnt','casual','registered'],axis=1).columns
target = df["cnt"].values
feature = df.drop(['cnt','casual','registered'],axis=1).values
from sklearn import preprocessing
# 对特征进行归一化等
feture = preprocessing.scale(feature)
In [13]:
print("name:", name)
print(feature)
print("target:", target)
name: Index(['instant', 'season', 'yr', 'mnth', 'hr', 'holiday', 'weekday',
'workingday', 'weathersit', 'temp', 'atemp', 'hum', 'windspeed'],
dtype='object')
[[1.0000e+00 1.0000e+00 0.0000e+00 ... 2.8790e-01 8.1000e-01 0.0000e+00]
[2.0000e+00 1.0000e+00 0.0000e+00 ... 2.7270e-01 8.0000e-01 0.0000e+00]
[3.0000e+00 1.0000e+00 0.0000e+00 ... 2.7270e-01 8.0000e-01 0.0000e+00]
...
[1.7377e+04 1.0000e+00 1.0000e+00 ... 2.5760e-01 6.0000e-01 1.6420e-01]
[1.7378e+04 1.0000e+00 1.0000e+00 ... 2.7270e-01 5.6000e-01 1.3430e-01]
[1.7379e+04 1.0000e+00 1.0000e+00 ... 2.7270e-01 6.5000e-01 1.3430e-01]]
target: [16 40 32 ... 90 61 49]
In [14]:
from sklearn.model_selection import ShuffleSplit
from sklearn.model_selection import learning_curve
In [15]:
# 高斯朴素贝叶斯估计器
from sklearn.naive_bayes import GaussianNB
cv = ShuffleSplit(n_splits=10, test_size=0.25, random_state=0)
estimator = GaussianNB()
X = feature; y= target
train_sizes, train_scores, test_scores = learning_curve( \
estimator, X, y, cv=cv, n_jobs=4)
train_scores_mean = np.mean(train_scores, axis=1)
train_scores_std = np.std(train_scores, axis=1)
test_scores_mean = np.mean(test_scores, axis=1)
test_scores_std = np.std(test_scores, axis=1)
In [16]:
print(train_scores_mean)
print(train_scores_std)
print(test_scores_mean)
print(test_scores_std)
print(train_sizes)
[0.27336915 0.0868508 0.05528739 0.04253044 0.03589842] [0.01112578 0.00475373 0.00310955 0.00220448 0.00113147] [0.00720368 0.0118527 0.0136939 0.01440736 0.01493671] [0.00143008 0.00249211 0.00216206 0.00155483 0.00156586] [ 1303 4236 7168 10101 13034]
In [17]:
plt.grid()
plt.fill_between(train_sizes, train_scores_mean - train_scores_std,
train_scores_mean + train_scores_std, alpha=0.1, color="r")
plt.fill_between(train_sizes, test_scores_mean - test_scores_std,
test_scores_mean + test_scores_std, alpha=0.1, color="g")
plt.plot(train_sizes, train_scores_mean, 'o-', color="r",
label="Training score")
plt.plot(train_sizes, test_scores_mean, 'o-', color="g",
label="Cross-validation score")
plt.legend(loc="best")
plt.show()
# 高斯贝叶斯朴素估计模型
In [53]:
def estm(estimator, X, y, n_jobs=1):
cv = ShuffleSplit(n_splits=10, test_size=0.25, random_state=0)
train_sizes, train_scores, test_scores = learning_curve( \
estimator, X, y, shuffle=True, cv=cv, n_jobs=n_jobs)
train_scores_mean = np.mean(train_scores, axis=1)
train_scores_std = np.std(train_scores, axis=1)
test_scores_mean = np.mean(test_scores, axis=1)
test_scores_std = np.std(test_scores, axis=1)
print("train_mean:", train_scores_mean)
print("test_mean:", test_scores_mean)
plt.grid()
plt.fill_between(train_sizes, train_scores_mean - train_scores_std,
train_scores_mean + train_scores_std, alpha=0.1, color="r")
plt.fill_between(train_sizes, test_scores_mean - test_scores_std,
test_scores_mean + test_scores_std, alpha=0.1, color="g")
plt.plot(train_sizes, train_scores_mean, 'o-', color="r",
label="Training score")
plt.plot(train_sizes, test_scores_mean, 'o-', color="g",
label="Cross-validation score")
plt.xlabel("Training examples")
plt.ylabel("Score")
plt.legend(loc="best")
plt.show()
In [54]:
# 岭回归估计器
from sklearn.linear_model import Ridge
ridge = Ridge(alpha=1, fit_intercept=True)
estm(estimator = ridge, X = feature, y= target) # 岭回归
train_mean: [0.39290346 0.3910804 0.38899554 0.39035204 0.38907019] test_mean: [0.38332278 0.38627641 0.38702503 0.38727722 0.38735053]
In [55]:
# svm估计器
from sklearn.svm import SVR
svm_est = SVR(gamma="scale")
estm(estimator = svm_est, X = feature, y= target)
train_mean: [-0.02780901 0.01771899 0.04605179 0.06373237 0.07625003] test_mean: [-0.03268469 0.01047544 0.03902897 0.05788598 0.07044845]
In [62]:
#跑上面的svm模型真的很慢
In [63]:
# 随机森林估计器
from sklearn.ensemble import RandomForestRegressor
rfr = RandomForestRegressor(n_estimators=100)
In [66]:
rfr_best = RandomForestRegressor(n_estimators=100, max_depth=10)
estm(rfr_best, X = feature, y= target)
train_mean: [0.95935031 0.95542485 0.95032115 0.94648625 0.94422042] test_mean: [0.85103462 0.90982391 0.91819348 0.92103033 0.92347536]
In [67]:
estm(rfr, X = feature, y= target)
train_mean: [0.98167846 0.98956945 0.99139242 0.99227577 0.99268494] test_mean: [0.86680905 0.92746839 0.93951654 0.94448014 0.9484968 ]
In [68]:
df2 = df_train.drop(['holiday', 'weekday','workingday',
'dteday','casual','registered'],axis=1)
# 删掉三个可能不相关的特征 :holiday, weekday, workingday
In [69]:
df2.head(5)
Out[69]:
| instant | season | yr | mnth | hr | weathersit | temp | atemp | hum | windspeed | cnt | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1 | 1 | 0 | 1 | 0 | 1 | 0.24 | 0.2879 | 0.81 | 0.0 | 16 |
| 1 | 2 | 1 | 0 | 1 | 1 | 1 | 0.22 | 0.2727 | 0.80 | 0.0 | 40 |
| 2 | 3 | 1 | 0 | 1 | 2 | 1 | 0.22 | 0.2727 | 0.80 | 0.0 | 32 |
| 3 | 4 | 1 | 0 | 1 | 3 | 1 | 0.24 | 0.2879 | 0.75 | 0.0 | 13 |
| 4 | 5 | 1 | 0 | 1 | 4 | 1 | 0.24 | 0.2879 | 0.75 | 0.0 | 1 |
In [75]:
# 减少三个特征后的随机森林效果
target2 = df2["cnt"].values
feature2 = df2.drop(['cnt'],axis=1).values
estm(rfr, X = feature2, y= target2)
train_mean: [0.96254768 0.96800351 0.96966206 0.97165311 0.97301559] test_mean: [0.73635969 0.76869966 0.78680719 0.79770389 0.80788745]
In [76]:
# 效果变差了，说明并不是冗余的特征
In [77]:
# 网格搜索法来寻找最优参数
from sklearn.model_selection import train_test_split
from sklearn.model_selection import GridSearchCV
X_train,X_test,y_train,y_test = train_test_split(feature, target, test_size=0.2,\
shuffle=True, random_state=None)
#构建网格参数
param_grid = {
'n_estimators': [10, 100, 500],
'max_depth': list(range(2,11))
}
#初始化模型
forest = RandomForestRegressor()
#初始化网格搜索
grid_search = GridSearchCV(estimator=forest, param_grid=param_grid, cv=5,
n_jobs=-1, verbose=1)
grid_search.fit(X_train, y_train)
#查看最好的参数选择
print(grid_search.best_params_)
#使用网格搜索得到的最好的参数选择进行模型训练
best_forest = grid_search.best_estimator_
best_forest.fit(X_train, y_train)
Fitting 5 folds for each of 27 candidates, totalling 135 fits
[Parallel(n_jobs=-1)]: Using backend LokyBackend with 4 concurrent workers. [Parallel(n_jobs=-1)]: Done 50 tasks | elapsed: 49.6s [Parallel(n_jobs=-1)]: Done 135 out of 135 | elapsed: 4.4min finished
{'max_depth': 10, 'n_estimators': 500}
Out[77]:
RandomForestRegressor(bootstrap=True, criterion='mse', max_depth=10,
max_features='auto', max_leaf_nodes=None,
min_impurity_decrease=0.0, min_impurity_split=None,
min_samples_leaf=1, min_samples_split=2,
min_weight_fraction_leaf=0.0, n_estimators=500, n_jobs=None,
oob_score=False, random_state=None, verbose=0, warm_start=False)
In [78]:
# for params, mean_score, scores in grid_search.grid_
print(grid_search.cv_results_)
{'mean_fit_time': array([ 0.13319407, 1.45894413, 7.69540601, 0.21543045, 2.12760754,
10.19142137, 0.25360661, 2.50598874, 12.50833755, 0.32933621,
3.05163474, 15.49317541, 0.40856681, 3.53402071, 17.51183238,
0.4156352 , 3.97993717, 19.81730919, 0.41597514, 4.47582984,
22.97348342, 0.52254672, 5.08494382, 25.21887112, 0.56973372,
5.89285698, 24.89119835]), 'std_fit_time': array([0.00597661, 0.06011933, 0.15387387, 0.06057044, 0.18664663,
0.35262105, 0.05621877, 0.0846163 , 0.12028449, 0.03099818,
0.16301284, 0.3957759 , 0.05232937, 0.18943767, 0.23094902,
0.05952818, 0.08112081, 0.35501047, 0.02914394, 0.25369788,
0.60628481, 0.07118387, 0.21263516, 0.33981729, 0.08466858,
0.14657052, 4.06957062]), 'mean_score_time': array([0.00348597, 0.02038913, 0.11069765, 0.00612483, 0.03590779,
0.14690261, 0.00495086, 0.03047194, 0.13486309, 0.00470958,
0.02950215, 0.1779366 , 0.00487576, 0.03642316, 0.21307054,
0.00582633, 0.04934931, 0.18337116, 0.00567908, 0.05458946,
0.25079675, 0.00578904, 0.0800024 , 0.26320786, 0.01405897,
0.05956931, 0.2289319 ]), 'std_score_time': array([0.00015086, 0.0009698 , 0.02113325, 0.00316835, 0.02203837,
0.03937412, 0.00127093, 0.0046392 , 0.00709288, 0.00045844,
0.00120708, 0.04808674, 0.00038644, 0.00786648, 0.0458313 ,
0.00170539, 0.01187062, 0.01696695, 0.0008307 , 0.02729879,
0.06539335, 0.00010715, 0.04716663, 0.05540412, 0.00835431,
0.0166167 , 0.03914065]), 'param_max_depth': masked_array(data=[2, 2, 2, 3, 3, 3, 4, 4, 4, 5, 5, 5, 6, 6, 6, 7, 7, 7,
8, 8, 8, 9, 9, 9, 10, 10, 10],
mask=[False, False, False, False, False, False, False, False,
False, False, False, False, False, False, False, False,
False, False, False, False, False, False, False, False,
False, False, False],
fill_value='?',
dtype=object), 'param_n_estimators': masked_array(data=[10, 100, 500, 10, 100, 500, 10, 100, 500, 10, 100, 500,
10, 100, 500, 10, 100, 500, 10, 100, 500, 10, 100, 500,
10, 100, 500],
mask=[False, False, False, False, False, False, False, False,
False, False, False, False, False, False, False, False,
False, False, False, False, False, False, False, False,
False, False, False],
fill_value='?',
dtype=object), 'params': [{'max_depth': 2, 'n_estimators': 10}, {'max_depth': 2, 'n_estimators': 100}, {'max_depth': 2, 'n_estimators': 500}, {'max_depth': 3, 'n_estimators': 10}, {'max_depth': 3, 'n_estimators': 100}, {'max_depth': 3, 'n_estimators': 500}, {'max_depth': 4, 'n_estimators': 10}, {'max_depth': 4, 'n_estimators': 100}, {'max_depth': 4, 'n_estimators': 500}, {'max_depth': 5, 'n_estimators': 10}, {'max_depth': 5, 'n_estimators': 100}, {'max_depth': 5, 'n_estimators': 500}, {'max_depth': 6, 'n_estimators': 10}, {'max_depth': 6, 'n_estimators': 100}, {'max_depth': 6, 'n_estimators': 500}, {'max_depth': 7, 'n_estimators': 10}, {'max_depth': 7, 'n_estimators': 100}, {'max_depth': 7, 'n_estimators': 500}, {'max_depth': 8, 'n_estimators': 10}, {'max_depth': 8, 'n_estimators': 100}, {'max_depth': 8, 'n_estimators': 500}, {'max_depth': 9, 'n_estimators': 10}, {'max_depth': 9, 'n_estimators': 100}, {'max_depth': 9, 'n_estimators': 500}, {'max_depth': 10, 'n_estimators': 10}, {'max_depth': 10, 'n_estimators': 100}, {'max_depth': 10, 'n_estimators': 500}], 'split0_test_score': array([0.4265695 , 0.43080089, 0.43333664, 0.52482063, 0.52868847,
0.52751703, 0.60295818, 0.60333284, 0.60722655, 0.6880848 ,
0.68472585, 0.68671101, 0.74101723, 0.73300685, 0.73615953,
0.83156864, 0.8328769 , 0.83403787, 0.88214921, 0.88438457,
0.88499887, 0.90604236, 0.9127264 , 0.9126119 , 0.92452809,
0.92869364, 0.92876554]), 'split1_test_score': array([0.43568706, 0.44799795, 0.44385583, 0.52410611, 0.52709677,
0.52811793, 0.5953118 , 0.60109326, 0.60361272, 0.66934986,
0.66205907, 0.66365064, 0.69897809, 0.70714238, 0.70854428,
0.79903639, 0.82000209, 0.82164397, 0.8671429 , 0.87276111,
0.87276547, 0.89823378, 0.90009964, 0.90172258, 0.91248584,
0.91660188, 0.91748047]), 'split2_test_score': array([0.43321726, 0.42151468, 0.42187655, 0.51774011, 0.51208051,
0.51265898, 0.57885442, 0.58301569, 0.58369676, 0.65188192,
0.65272815, 0.6526802 , 0.70556499, 0.70892637, 0.71218678,
0.80873436, 0.81543464, 0.81552178, 0.86679453, 0.86753016,
0.87116847, 0.90064449, 0.90381184, 0.90346152, 0.9161736 ,
0.91912169, 0.92073253]), 'split3_test_score': array([0.44929382, 0.43668538, 0.43876207, 0.53115374, 0.53166857,
0.53366559, 0.58874033, 0.59201368, 0.59359413, 0.66397063,
0.66445918, 0.66281597, 0.71152393, 0.71379972, 0.71702243,
0.81702849, 0.81725307, 0.8207766 , 0.87849206, 0.88226444,
0.88122158, 0.89888095, 0.91028826, 0.91049059, 0.92404634,
0.9293586 , 0.92856867]), 'split4_test_score': array([0.40932677, 0.40906575, 0.40999022, 0.51882518, 0.51824527,
0.51787675, 0.56934065, 0.56913182, 0.57026932, 0.63981228,
0.6448662 , 0.64454813, 0.69009841, 0.69197979, 0.6922864 ,
0.80578953, 0.80934544, 0.80937264, 0.85942446, 0.86248597,
0.86060782, 0.89261474, 0.89356756, 0.89303781, 0.90865972,
0.91347295, 0.91428701]), 'mean_test_score': array([0.4308191 , 0.42921384, 0.42956501, 0.52332891, 0.52355572,
0.523967 , 0.58704223, 0.58971877, 0.5916813 , 0.66262144,
0.66176871, 0.6620824 , 0.70943777, 0.71097219, 0.71324112,
0.81243163, 0.81898325, 0.82027132, 0.8708009 , 0.87388547,
0.87415291, 0.89928377, 0.90409905, 0.90426524, 0.91717884,
0.92144976, 0.92196692]), 'std_test_score': array([0.01304282, 0.01323909, 0.01220871, 0.00480755, 0.00727445,
0.00760069, 0.01187401, 0.01255947, 0.01348972, 0.01631609,
0.01342985, 0.01418063, 0.01732113, 0.01321422, 0.01415489,
0.01117714, 0.00777887, 0.0081609 , 0.00832797, 0.0083907 ,
0.0085073 , 0.00431899, 0.00692399, 0.00695073, 0.00627327,
0.00644297, 0.00583823]), 'rank_test_score': array([25, 27, 26, 24, 23, 22, 21, 20, 19, 16, 18, 17, 15, 14, 13, 12, 11,
10, 9, 8, 7, 6, 5, 4, 3, 2, 1], dtype=int32), 'split0_train_score': array([0.42601463, 0.42946586, 0.43192614, 0.52110995, 0.52461488,
0.5237484 , 0.58845221, 0.58731621, 0.59194356, 0.6618245 ,
0.65969967, 0.66122801, 0.72539732, 0.71564364, 0.71902706,
0.82535444, 0.82712128, 0.82870432, 0.88415985, 0.88549359,
0.88635991, 0.91569962, 0.92103971, 0.92095331, 0.94275207,
0.94450779, 0.94449826]), 'split1_train_score': array([0.42398261, 0.4378589 , 0.43313634, 0.52467323, 0.52697513,
0.52796217, 0.58939228, 0.59543514, 0.59832749, 0.67583265,
0.66969714, 0.67155511, 0.71801122, 0.72443821, 0.72567976,
0.81395794, 0.83166674, 0.83248985, 0.87988359, 0.88652003,
0.88689117, 0.91690117, 0.91990356, 0.92075336, 0.93930202,
0.94386384, 0.94413779]), 'split2_train_score': array([0.44351265, 0.43157532, 0.43184905, 0.53441443, 0.52803298,
0.52833472, 0.58927397, 0.59361595, 0.59437662, 0.66204444,
0.66184514, 0.66157483, 0.70696482, 0.71097096, 0.71479343,
0.81751595, 0.82773331, 0.82808917, 0.8811396 , 0.88224729,
0.88442762, 0.91551727, 0.92007969, 0.92003873, 0.94123196,
0.94369118, 0.94453403]), 'split3_train_score': array([0.438328 , 0.42591528, 0.42791695, 0.52071378, 0.52113498,
0.52323859, 0.5881326 , 0.59101744, 0.59260369, 0.66530009,
0.66624063, 0.66447488, 0.71097453, 0.71388843, 0.71665822,
0.81099975, 0.81496985, 0.81856281, 0.88168968, 0.88491129,
0.8837703 , 0.91048218, 0.91870837, 0.91917717, 0.93861616,
0.94248367, 0.94236033]), 'split4_train_score': array([0.43071721, 0.43098549, 0.43152661, 0.52696716, 0.52576168,
0.5256326 , 0.59580671, 0.59585888, 0.59667716, 0.67613408,
0.6768601 , 0.67622093, 0.729107 , 0.73067035, 0.73081437,
0.83574275, 0.83807291, 0.83837898, 0.88949672, 0.89255113,
0.89160827, 0.92314265, 0.92437289, 0.92455876, 0.94181603,
0.94628481, 0.9471218 ]), 'mean_train_score': array([0.43251102, 0.43116017, 0.43127102, 0.52557571, 0.52530393,
0.5257833 , 0.59021155, 0.59264872, 0.5947857 , 0.66822715,
0.66686854, 0.66701075, 0.71809098, 0.71912232, 0.72139457,
0.82071417, 0.82791282, 0.82924503, 0.88327389, 0.88634467,
0.88661145, 0.91634858, 0.92082084, 0.92109627, 0.94074365,
0.94416626, 0.94453044]), 'std_train_score': array([0.00738956, 0.00388473, 0.00176396, 0.00498909, 0.00237929,
0.00209267, 0.00283807, 0.00316549, 0.00241363, 0.00645214,
0.00608036, 0.00591437, 0.00835676, 0.00731512, 0.00597844,
0.0089178 , 0.00755644, 0.00647605, 0.00340786, 0.00340935,
0.00275503, 0.00404948, 0.00192448, 0.00184001, 0.00155092,
0.00124524, 0.00152182])}
In [79]:
rfr_best = RandomForestRegressor(n_estimators=100, max_depth=10)
estm(rfr_best, X = feature, y= target)
train_mean: [0.96268762 0.95542684 0.95017855 0.94705958 0.94422869] test_mean: [0.85949261 0.91012582 0.91893709 0.92240978 0.92330138]
In [80]:
# 尝试原来博客中的参数，使用五折交叉验证而不是shufflesplit，树的深度不设置上限
In [82]:
def estm2(estimator, X, y, n_jobs=1,cv = 5):
train_sizes, train_scores, test_scores = learning_curve( \
estimator, X, y, shuffle=True, cv=cv, n_jobs=n_jobs)
train_scores_mean = np.mean(train_scores, axis=1)
train_scores_std = np.std(train_scores, axis=1)
test_scores_mean = np.mean(test_scores, axis=1)
test_scores_std = np.std(test_scores, axis=1)
print("train_mean:", train_scores_mean)
print("test_mean:", test_scores_mean)
plt.grid()
plt.fill_between(train_sizes, train_scores_mean - train_scores_std,
train_scores_mean + train_scores_std, alpha=0.1, color="r")
plt.fill_between(train_sizes, test_scores_mean - test_scores_std,
test_scores_mean + test_scores_std, alpha=0.1, color="g")
plt.plot(train_sizes, train_scores_mean, 'o-', color="r",
label="Training score")
plt.plot(train_sizes, test_scores_mean, 'o-', color="g",
label="Cross-validation score")
plt.xlabel("Training examples")
plt.ylabel("Score")
plt.legend(loc="best")
plt.show()
In [83]:
r = RandomForestRegressor(n_estimators=100)
estm2(r, X = feature, y= target)
train_mean: [0.98351078 0.99064278 0.99200975 0.99274763 0.99327549] test_mean: [0.701878 0.75343972 0.77403515 0.79018347 0.79756075]
In [84]:
# 通过还原数据得到结论，树的深度剪枝可以避免过拟合，用shufflesplit比5折交叉验证的效果要提升很多。
In [ ]:
In [93]:
rfr_best = RandomForestRegressor(n_estimators=100,max_depth=10).fit(feature, target)
test_scale = feature
pred = np.array(rfr_best.predict(test_scale))
dic = dict()
dic.update({'pred_result':pred, "datatime":df_train["dteday"], "hour":df_train["hr"]})
import pandas
df_pre_result = pandas.DataFrame(data=dic)
df_pre_result.head(15)
Out[93]:
| pred_result | datatime | hour | |
|---|---|---|---|
| 0 | 25.197721 | 2011-01-01 | 0 |
| 1 | 28.853542 | 2011-01-01 | 1 |
| 2 | 25.540445 | 2011-01-01 | 2 |
| 3 | 10.315996 | 2011-01-01 | 3 |
| 4 | 2.496642 | 2011-01-01 | 4 |
| 5 | 2.477209 | 2011-01-01 | 5 |
| 6 | 3.044991 | 2011-01-01 | 6 |
| 7 | 30.548464 | 2011-01-01 | 7 |
| 8 | 43.904548 | 2011-01-01 | 8 |
| 9 | 51.795345 | 2011-01-01 | 9 |
| 10 | 69.528911 | 2011-01-01 | 10 |
| 11 | 71.463207 | 2011-01-01 | 11 |
| 12 | 92.183035 | 2011-01-01 | 12 |
| 13 | 111.922991 | 2011-01-01 | 13 |
| 14 | 113.988334 | 2011-01-01 | 14 |
In [ ]:
In [ ]: