随机森林算法python实现

瞎BB

1.实现根据样本数据（用眼距离distance、最长持续用眼时长duration、总用眼时长total_time、户外运动时长outdoor、用眼角度angle、健康环境光照用眼比例proportion）判别是否需要近视预警
2.样本实在太少，结果还行，原理都是一样的

代码

导入数据

import pandas
patients = pandas.read_csv("data.csv")
patients.head(5)

切分训练集测试集

from sklearn.model_selection import train_test_split
patients_data=patients.loc[:,'distance':'proportion']
patients_target=patients.loc[:,'warning']
data_train,data_test,target_train,target_test=train_test_split(patients_data,patients_target,test_size=0.1,random_state=42)

找到最有用的几个属性

import numpy as np
from sklearn.feature_selection import SelectKBest, f_classif
import matplotlib.pyplot as plt
predictors = ["distance", "duration", "total_time", "outdoor", "angle", "proportion"]

selector = SelectKBest(f_classif, k=5)
selector.fit(data_train, target_train)

scores = -np.log10(selector.pvalues_)

plt.bar(range(len(predictors)), scores)
plt.xticks(range(len(predictors)), predictors, rotation='vertical')
plt.show()

根据上面的代码更改属性

predictors_best = ["distance", "total_time", "angle", "proportion"]
data_train = data_train[predictors_best]
data_test = data_test[predictors_best]

参数组合遍历找最优

from sklearn.model_selection import GridSearchCV
tree_param_grid = { 'min_samples_split': list((2,3,4)),'n_estimators':list((3,5,10,15,20,25,30,35,40,45,50))}
grid = GridSearchCV(RandomForestClassifier(),param_grid=tree_param_grid, cv=kf)#(算法，调节参数（用字典形式），交叉验证次数)
grid.fit(data_train, target_train)#训练集
grid.cv_results_ , grid.best_params_, grid.best_score_#得分，最优参数，最优得分

随机森林

from sklearn import model_selection
from sklearn.ensemble import RandomForestClassifier
rf = RandomForestClassifier(random_state=1, n_estimators=35, min_samples_split=2, min_samples_leaf=2)
#交叉验证
kf = model_selection.KFold(n_splits=3)
scores = model_selection.cross_val_score(rf, data_train, target_train, cv=kf)
print(scores.mean())

样本数据

sample	distance	duration	total_time	outdoor	angle	proportion	warning(1 yes 0 no)
1	20	72	344	148	11	81	1
2	34	68	263	135	7	50	1
3	25	98	357	32	12	64	1
4	37	65	291	157	8	89	0
5	34	151	162	169	18	63	1
6	30	178	259	146	32	50	1
7	20	35	134	37	23	68	0
8	39	111	169	87	4	52	0
9	22	44	265	136	14	76	1
10	39	151	219	140	2	55	0
11	21	179	184	64	18	60	1
12	25	41	241	71	16	72	1
13	18	171	286	131	35	89	1
14	32	33	236	102	29	50	1
15	20	133	226	124	17	81	1
16	17	148	236	66	32	75	1
17	34	111	214	57	5	88	0
18	24	85	163	155	14	32	1
19	32	165	276	146	33	52	1
20	25	124	359	171	33	70	0
21	31	51	167	47	25	47	0
22	31	63	352	58	22	44	1
23	16	58	164	45	13	73	0
24	29	37	326	104	33	68	1
25	34	47	197	59	5	66	0
26	36	123	185	165	26	70	0
27	25	126	171	45	23	33	1
28	31	84	98	37	30	51	1
29	30	92	153	114	14	48	0
30	29	178	278	146	27	45	1