随机森林的应用简例

用jupyter lab 做的 太多了懒得排版

%matplotlib inline
import numpy as np
import pandas as pd
from sklearn.model_selection import GridSearchCV
from sklearn.ensemble import RandomForestClassifier
df_t = pd.read_excel(r'D:\EdgeDownloadPlace\复赛数据集\train.xlsx',header=None)
​
import re
with open(r'D:\EdgeDownloadPlace\复赛数据集\features.txt') as f:
    features = re.findall('[0-9] (.*)\n', f.read())
features.insert(0,'uid')
features.append('target')
df_t.columns = features
​
df_t = df_t.drop(columns = 'uid')
arr_t = df_t.values
import time
start_time = time.time()
param_grid = {'n_estimators' : np.arange(1,201,40)}
rfc = RandomForestClassifier(random_state = 435681971
                            ,criterion = 'entropy')
gs = GridSearchCV(rfc, param_grid, cv=4)
gs.fit(arr_t[:,:-1],arr_t[:,-1])
​
peak_n_lst = [gs.best_score_, gs.best_params_]
end_time = time.time()
time_span = end_time - start_time
peak_n_lst
​
[0.9373961218836566, {'n_estimators': 161}]
time_span
107.33542943000793
peak_n = peak_n_lst[1]['n_estimators']
start_time = time.time()
param_grid = {'n_estimators' : np.arange(peak_n-20,peak_n+20)}
rfc = RandomForestClassifier(random_state = 435681971
                            ,criterion = 'entropy')
gs = GridSearchCV(rfc, param_grid, cv=4)
gs.fit(arr_t[:,:-1],arr_t[:,-1])
​
peak_n_lst = [gs.best_score_, gs.best_params_]
end_time = time.time()
time_span = end_time - start_time
peak_n = peak_n_lst[1]['n_estimators']
peak_n
170
start_time = time.time()
param_grid = {'max_depth' : np.arange(1,561//2,30)}
rfc = RandomForestClassifier(random_state = 435681971
                            ,n_estimators = peak_n
                            ,criterion = 'entropy')
gs = GridSearchCV(rfc, param_grid, cv=4)
gs.fit(arr_t[:,:-1],arr_t[:,-1])
​
peak_depth_lst = [gs.best_score_, gs.best_params_]
end_time = time.time()
time_span = end_time - start_time
peak_depth_lst
[0.9386426592797784, {'max_depth': 31}]
time_span
407.307009935379
peak_depth = peak_depth_lst[1]['max_depth']
peak_depth
31
start_time = time.time()
param_grid = {'max_depth' : np.arange(peak_depth-20, peak_depth+30)}
rfc = RandomForestClassifier(random_state = 435681971
                            ,n_estimators = peak_n
                            ,criterion = 'entropy')
gs = GridSearchCV(rfc, param_grid, cv=4)
gs.fit(arr_t[:,:-1],arr_t[:,-1])
​
peak_depth_lst = [gs.best_score_, gs.best_params_]
end_time = time.time()
time_span = end_time - start_time
peak_depth_lst
[0.9393351800554016, {'max_depth': 14}]
time_span/60
​
31.484332279364267
peak_depth = peak_depth_lst[1]['max_depth']
peak_depth
14
start_time = time.time()
param_grid = {'min_samples_split' : np.arange(2,125,30)}
rfc = RandomForestClassifier(random_state = 435681971
                            ,n_estimators = peak_n
                            ,max_depth = peak_depth
                            ,criterion = 'entropy')
gs = GridSearchCV(rfc, param_grid, cv=4)
gs.fit(arr_t[:,:-1],arr_t[:,-1])
​
peak_depth_lst = [gs.best_score_, gs.best_params_]
end_time = time.time()
time_span = end_time - start_time
time_span/60
3.157407291730245
#peak_depth_lst 上面忘记换名字了
peak_minss = peak_depth_lst[1]['min_samples_split']
peak_minss
2
start_time = time.time()
param_grid = {'min_samples_split' : np.arange(2,30)}
rfc = RandomForestClassifier(random_state = 435681971
                            ,n_estimators = peak_n
                            ,max_depth = peak_depth
                            ,criterion = 'entropy')
gs = GridSearchCV(rfc, param_grid, cv=4)
gs.fit(arr_t[:,:-1],arr_t[:,-1])
​
peak_depth_lst = [gs.best_score_, gs.best_params_]
end_time = time.time()
time_span = end_time - start_time
time_span
1053.7646670341492
peak_depth_lst
[0.9393351800554016, {'min_samples_split': 2}]
peak_minss = peak_depth_lst[1]['min_samples_split']
peak_minss
2
import matplotlib.pyplot as plt
peak_n = 170
peak_depth = 14
peak_minss = 3
#rfc = RandomForestClassifier(random_state = 435681971
#                            ,n_estimators = peak_n
#                            ,max_depth = peak_depth
#                            ,min_samples_split = peak_minss
#                            ,oob_score = True)
#rfc.fit(arr_t[:,:-1],arr_t[:,-1])
#rfc.oob_score_
#plt.figure(figsize = [20,5])
​
#score_lst=[]
#for i in range(30):
#    rfc = RandomForestClassifier(#random_state = 435681971
#                                n_estimators = peak_n
#                                ,max_depth = peak_depth
#                                ,min_samples_split = peak_minss
#                                ,oob_score = True)
#    rfc.fit(arr_t[:,:-1],arr_t[:,-1])
#    score_lst.append(rfc.oob_score_)
#plt.plot(range(1,31),score_lst)
​
score_lst=[]
for i in range(30):
    rfc = RandomForestClassifier(#random_state = 435681971
                                n_estimators = peak_n
                                ,max_depth = peak_depth
                                ,min_samples_split = peak_minss
                                ,oob_score = True
                                ,criterion = 'entropy')
    rfc.fit(arr_t[:,:-1],arr_t[:,-1])
    score_lst.append(rfc.oob_score_)
plt.plot(range(1,31),score_lst,color = 'red')
​
plt.show()   

plt.figure(figsize = [15,6])
plt.plot(range(1,31),score_lst,color = 'red')
plt.show()

while True:
    rfc = RandomForestClassifier(#random_state = 435681971
                                n_estimators = peak_n
                                ,max_depth = peak_depth
                                ,min_samples_split = peak_minss
                                ,oob_score = True
                                ,criterion = 'entropy')
    rfc.fit(arr_t[:,:-1],arr_t[:,-1])
    if rfc.oob_score_ > 0.978:
        break
df_a = pd.read_excel(r'D:\EdgeDownloadPlace\复赛数据集\test.xlsx',header=None)
​
df_a.columns = features[:-1]
​
df_a = df_a.drop(columns= 'uid')
df_a
tBodyAcc-mean()-X	tBodyAcc-mean()-Y	tBodyAcc-mean()-Z	tBodyAcc-std()-X	tBodyAcc-std()-Y	tBodyAcc-std()-Z	tBodyAcc-mad()-X	tBodyAcc-mad()-Y	tBodyAcc-mad()-Z	tBodyAcc-max()-X	...	fBodyBodyGyroJerkMag-meanFreq()	fBodyBodyGyroJerkMag-skewness()	fBodyBodyGyroJerkMag-kurtosis()	angle(tBodyAccMean,gravity)	angle(tBodyAccJerkMean),gravityMean)	angle(tBodyGyroMean,gravityMean)	angle(tBodyGyroJerkMean,gravityMean)	angle(X,gravityMean)	angle(Y,gravityMean)	angle(Z,gravityMean)
0	0.278	-0.01640	-0.1240	-0.998	-0.9750	-0.960	-0.999	-0.9750	-0.958	-0.9430	...	0.1580	-0.5950	-0.861	0.0535	-0.00743	-0.733	0.7040	-0.845	0.180	-0.0543
1	0.281	-0.00996	-0.1060	-0.995	-0.9730	-0.986	-0.995	-0.9740	-0.986	-0.9400	...	0.2670	0.3400	0.140	-0.0206	-0.12800	-0.483	-0.0707	-0.848	0.190	-0.0344
2	0.277	-0.01470	-0.1070	-0.999	-0.9910	-0.993	-0.999	-0.9910	-0.992	-0.9430	...	0.7400	-0.5640	-0.766	0.1060	-0.09030	-0.132	0.4990	-0.850	0.189	-0.0351
3	0.279	-0.02300	-0.1220	-0.997	-0.9750	-0.983	-0.997	-0.9730	-0.984	-0.9420	...	0.6620	-0.7820	-0.954	-0.1220	-0.02910	-0.013	-0.0569	-0.761	0.263	0.0242
4	0.280	-0.01390	-0.1060	-0.998	-0.9880	-0.990	-0.998	-0.9880	-0.992	-0.9420	...	0.4290	-0.3290	-0.597	-0.0283	0.09240	-0.822	0.3680	-0.759	0.264	0.0297
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
3074	0.231	-0.04230	-0.0899	-0.309	-0.0791	-0.152	-0.391	-0.0870	-0.257	0.0562	...	-0.0310	-0.1390	-0.589	0.2730	0.85600	-0.962	0.9530	-0.657	0.276	0.1770
3075	0.357	-0.04460	-0.1300	-0.314	-0.0556	-0.173	-0.386	-0.0575	-0.217	0.0262	...	0.0168	-0.1630	-0.593	-0.7110	-0.06120	-0.706	0.0646	-0.660	0.274	0.1760
3076	0.284	-0.00796	-0.1190	-0.309	-0.0804	-0.211	-0.369	-0.0971	-0.301	-0.1170	...	-0.1100	0.0245	-0.393	-0.0761	-0.23900	0.960	0.0866	-0.657	0.272	0.1830
3077	0.207	0.02460	-0.1040	-0.365	-0.1690	-0.216	-0.449	-0.1860	-0.326	-0.1760	...	-0.2140	-0.3520	-0.734	0.5350	-0.25700	0.927	-0.0843	-0.657	0.267	0.1880
3078	0.331	-0.06400	-0.1170	-0.068	0.1560	-0.317	-0.149	0.0701	-0.291	0.4120	...	-0.0214	-0.0863	-0.468	-0.3510	-0.33600	0.967	-0.7150	-0.810	0.185	0.1210
3079 rows × 561 columns

arr_a = df_a.values
arr_a
array([[ 0.278  , -0.0164 , -0.124  , ..., -0.845  ,  0.18   , -0.0543 ],
       [ 0.281  , -0.00996, -0.106  , ..., -0.848  ,  0.19   , -0.0344 ],
       [ 0.277  , -0.0147 , -0.107  , ..., -0.85   ,  0.189  , -0.0351 ],
       ...,
       [ 0.284  , -0.00796, -0.119  , ..., -0.657  ,  0.272  ,  0.183  ],
       [ 0.207  ,  0.0246 , -0.104  , ..., -0.657  ,  0.267  ,  0.188  ],
       [ 0.331  , -0.064  , -0.117  , ..., -0.81   ,  0.185  ,  0.121  ]])
answer = rfc.predict(arr_a).astype(np.int8).tolist()
len(answer)
3079
answer_df = pd.DataFrame(answer)
answer_df
0
0	5
1	5
2	5
3	5
4	5
...	...
3074	2
3075	2
3076	2
3077	2
3078	3
3079 rows × 1 columns

answer_df.to_excel(r'D:\EdgeDownloadPlace\复赛数据集\ANS\20201104try.xlsx')
print('ok')
ok