​代码较为简单。

import numpy as np # linear algebraimport pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)import os

import matplotlib.pyplot as pltimport seaborn as sns%matplotlib inline​import warningswarnings.filterwarnings('ignore')

df=pd.read_csv("Battery_RUL.csv")df.head()

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 15064 entries, 0 to 15063
Data columns (total 9 columns):
 #   Column                     Non-Null Count  Dtype  
---  ------                     --------------  -----  
 0   Cycle_Index                15064 non-null  float64
 1   Discharge Time (s)         15064 non-null  float64
 2   Decrement 3.6-3.4V (s)     15064 non-null  float64
 3   Max. Voltage Dischar. (V)  15064 non-null  float64
 4   Min. Voltage Charg. (V)    15064 non-null  float64
 5   Time at 4.15V (s)          15064 non-null  float64
 6   Time constant current (s)  15064 non-null  float64
 7   Charging time (s)          15064 non-null  float64
 8   RUL                        15064 non-null  int64  
dtypes: float64(8), int64(1)
memory usage: 1.0 MB

df.describe()

numlist=df.columns fig = plt.figure(figsize=(20,20))for i in range(len(numlist)):    plt.subplot(3,3,i+1)    plt.title(numlist[i],fontsize=12)    plt.hist(df[numlist[i]])​plt.tight_layout()

sns.pairplot(df)

​

plt.figure(figsize = (15,8))sns.heatmap(df.corr(),annot=True, cbar=False, cmap='Blues', fmt='.1f')

df1=df.drop(['Cycle_Index','Discharge Time (s)', 'Decrement 3.6-3.4V (s)', 'Time constant current (s)','Charging time (s)'],axis=1)from pycaret.regression import *exp_reg = setup(df1, target = 'RUL',remove_outliers=True,session_id=1)

Description	Value
0	Session id	1
1	Target	RUL
2	Target type	Regression
3	Original data shape	(15064, 4)
4	Transformed data shape	(14536, 4)
5	Transformed train set shape	(10016, 4)
6	Transformed test set shape	(4520, 4)
7	Numeric features	3
8	Preprocess	True
9	Imputation type	simple
10	Numeric imputation	mean
11	Categorical imputation	mode
12	Remove outliers	True
13	Outliers threshold	0.050000
14	Fold Generator	KFold
15	Fold Number	10
16	CPU Jobs	-1
17	Use GPU	False
18	Log Experiment	False
19	Experiment Name	reg-default-name
20	USI	dc44

compare_models()

	Model	MAE	MSE	RMSE	R2	RMSLE	MAPE	TT (Sec)
et	Extra Trees Regressor	21.7428	5805.6217	75.5050	0.9441	0.3375	0.2090	0.1840
rf	Random Forest Regressor	23.3808	6218.2462	78.2757	0.9402	0.3703	0.2286	0.2520
xgboost	Extreme Gradient Boosting	28.1191	6649.0490	80.7887	0.9360	0.3634	0.2583	0.1810
lightgbm	Light Gradient Boosting Machine	33.3121	6980.3989	82.9109	0.9328	0.3918	0.3013	0.1180
dt	Decision Tree Regressor	24.7941	7224.9013	84.4653	0.9305	0.4137	0.2263	0.0450
gbr	Gradient Boosting Regressor	41.4292	7696.6475	87.1895	0.9259	0.4326	0.3767	0.1220
ada	AdaBoost Regressor	53.8692	8601.3932	92.3123	0.9172	0.5116	0.5761	0.0830
knn	K Neighbors Regressor	47.6369	8847.0397	93.7257	0.9149	0.5202	0.3359	0.0500
dummy	Dummy Regressor	278.8523	104050.8450	322.5486	-0.0008	1.0563	3.3094	0.0420
lr	Linear Regression	215.7580	3866029.7278	1866.2254	-36.1340	0.6161	0.7615	0.2940
lar	Least Angle Regression	215.7595	3866065.5042	1866.2348	-36.1343	0.6161	0.7614	0.0440
ridge	Ridge Regression	220.8640	4141461.0688	1930.9796	-38.7797	0.6138	0.7712	0.0450
par	Passive Aggressive Regressor	268.5288	4193895.0303	1907.7742	-39.3176	0.6897	0.9640	0.0470
br	Bayesian Ridge	232.6850	4716618.3836	2054.9328	-44.3122	0.6308	0.8136	0.0430
lasso	Lasso Regression	233.8883	4790127.0360	2073.9986	-45.0074	0.6325	0.8140	0.2320
llar	Lasso Least Angle Regression	233.8883	4790127.6230	2073.9987	-45.0074	0.6325	0.8140	0.0460
omp	Orthogonal Matching Pursuit	233.8889	4790152.8777	2074.0042	-45.0076	0.6324	0.8140	0.0340
en	Elastic Net	233.8890	4790175.9367	2074.0099	-45.0078	0.6325	0.8140	0.0420
huber	Huber Regressor	235.3448	4915415.1237	2086.1153	-46.1760	0.6331	0.8195	0.0460

ExtraTreesRegressor(n_jobs=-1, random_state=1)

reg_model = create_model('et')

plot_model(reg_model,plot='error')

plot_model(reg_model,plot='residuals')

plot_model(reg_model,plot='feature')

xgb = create_model('xgboost',verbose=False)rf = create_model('rf', verbose = False)blend_specific = blend_models(estimator_list = [rf,xgb,reg_model])

工学博士，担任《Mechanical System and Signal Processing》《中国电机工程学报》《控制与决策》等期刊审稿专家，擅长领域：现代信号处理，机器学习，深度学习，数字孪生，时间序列分析，设备缺陷检测、设备异常检测、设备智能故障诊断与健康管理PHM等。