Machine-learning-aided-design-of-aluminum-alloys复现

Al-Cu-Mg-x合金被选为本工作的基础材料

import numpy as np
import pandas as pd
import seaborn as sns
import matplotlib.pyplot as plt

data=pd.read_excel('1-s2.0-S2352492820329081-mmc1.xlsx',header=1)
data.head()

	Hardness	Al	Cu	Mg	Si	Zn	Zr	Mn	Ag	Li	...	atomic radius	Specfic heat	Heat of Vaporization	thermal conductivity	group	period	Time (min)	Temearture (K)	Unnamed: 34	Reference
0	169	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	60.0	403	NaN	Scr mat 64, 21
1	174	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	180.0	403	NaN	Scr mat 64, 21
2	180	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	360.0	403	NaN	Scr mat 64, 21
3	196	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	720.0	403	NaN	Scr mat 64, 21
4	187	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	900.0	403	NaN	Scr mat 64, 21

5 rows × 36 columns

data.dropna(axis=1,inplace=True)
data.head()

	Hardness	Al	Cu	Mg	Si	Zn	Zr	Mn	Ag	Li	...	Fusion heat	atomic radius	Specfic heat	Heat of Vaporization	thermal conductivity	group	period	Time (min)	Temearture (K)	Reference
0	169	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	60.0	403	Scr mat 64, 21
1	174	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	180.0	403	Scr mat 64, 21
2	180	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	360.0	403	Scr mat 64, 21
3	196	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	720.0	403	Scr mat 64, 21
4	187	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	900.0	403	Scr mat 64, 21

5 rows × 35 columns

data.reset_index(drop=True)
data.drop(axis=1,labels=data.columns[-1],inplace=True)
data.head()

	Hardness	Al	Cu	Mg	Si	Zn	Zr	Mn	Ag	Li	...	Electroaffinity	Fusion heat	atomic radius	Specfic heat	Heat of Vaporization	thermal conductivity	group	period	Time (min)	Temearture (K)
0	169	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	41.4625	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	60.0	403
1	174	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	41.4625	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	180.0	403
2	180	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	41.4625	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	360.0	403
3	196	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	41.4625	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	720.0	403
4	187	0.9338	0.015	0.025	0.0	0.057	0.0	0.0	0.0	0.0	...	41.4625	10.82461	124.0824	897.5312	288.0864	236.283	13.0384	3.1644	900.0	403

5 rows × 34 columns

各种 Al-Cu-Mg-x 基合金的成分和时效条件（温度和时间）以及相关性能（硬度）等特征

物理特性$X_i$：元素i的电负性，原子序数，原子质量，原子半径，价电子，沸点，比热，熔点，汽化热，聚变热，组，周期，电亲和力，密度，导热性

$C_i$:元素i在合金的浓度

$X=\Sigma C_i{X}_i$

X是特征加权分数

data.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1591 entries, 0 to 1590
Data columns (total 34 columns):
 #   Column                 Non-Null Count  Dtype  
---  ------                 --------------  -----  
 0   Hardness               1591 non-null   int64  
 1   Al                     1591 non-null   float64
 2   Cu                     1591 non-null   float64
 3   Mg                     1591 non-null   float64
 4   Si                     1591 non-null   float64
 5   Zn                     1591 non-null   float64
 6   Zr                     1591 non-null   float64
 7   Mn                     1591 non-null   float64
 8   Ag                     1591 non-null   float64
 9   Li                     1591 non-null   float64
 10  Ca                     1591 non-null   int64  
 11  Fe                     1591 non-null   float64
 12  Ti                     1591 non-null   float64
 13  Sn                     1591 non-null   float64
 14  Cr                     1591 non-null   float64
 15  Ge                     1591 non-null   float64
 16  Sc                     1591 non-null   float64
 17  EN                     1591 non-null   float64
 18  VE                     1591 non-null   float64
 19  Atomic number          1591 non-null   float64
 20  mass                   1591 non-null   float64
 21  Melting point          1591 non-null   float64
 22  Boiling point          1591 non-null   float64
 23  Density                1591 non-null   float64
 24  Electroaffinity        1591 non-null   float64
 25  Fusion heat            1591 non-null   float64
 26  atomic radius          1591 non-null   float64
 27  Specfic heat           1591 non-null   float64
 28  Heat of Vaporization   1591 non-null   float64
 29  thermal conductivity   1591 non-null   float64
 30  group                  1591 non-null   float64
 31  period                 1591 non-null   float64
 32  Time  (min)            1591 non-null   float64
 33  Temearture (K)         1591 non-null   int64  
dtypes: float64(31), int64(3)
memory usage: 422.7 KB

data.isnull().sum()

Hardness                 0
Al                       0
Cu                       0
Mg                       0
Si                       0
Zn                       0
Zr                       0
Mn                       0
Ag                       0
Li                       0
Ca                       0
Fe                       0
Ti                       0
Sn                       0
Cr                       0
Ge                       0
Sc                       0
EN                       0
VE                       0
Atomic number            0
mass                     0
Melting point            0
Boiling point            0
Density                  0
Electroaffinity          0
Fusion heat              0
atomic radius            0
Specfic heat             0
Heat of Vaporization     0
thermal conductivity     0
group                    0
period                   0
Time  (min)              0
Temearture (K)           0
dtype: int64

LOF异常值检验

每个样本的异常值称为局部异常因子。它测量给定样品相对于其邻近样品密度的局部偏差。它是局部的，因为异常评分取决于物体相对于周围邻居的隔离程度。更准确地说，局部性是由 k 个最近邻居给出的，它们的距离用来估计局部密度。通过比较样本的局部密度与其邻近样本的局部密度，我们可以识别出密度大大低于其邻近样本的样本。这些被认为是异常值。

• n_neighbors:如果 n _ neighbors大于提供的样本数，则将使用所有样本。
• algorithm:{‘auto’, ‘ball_tree’, ‘kd_tree’, ‘brute’}

df=data.values
df.shape

(1591, 34)

from sklearn.neighbors import LocalOutlierFactor

#10个样本一组，异常值比例为0.055(和论文差不多)
LOF=LocalOutlierFactor(n_neighbors=10,contamination=0.055)
#inliner为1，outliner为-1
lables=LOF.fit_predict(df)
#LOF检验完的数据
df=df[lables>0]
df.shape

(1503, 34)

y=df[:,0]
y.shape

(1503,)

X=df[:,1:]
X.shape

(1503, 33)

划分数据和标准化数据

from sklearn.model_selection import train_test_split
X_train,X_test,y_train,y_test=train_test_split(X,
                                               y,
                                               test_size=0.4,
                                               random_state=0)

X_train.shape,X_test.shape

((901, 33), (602, 33))

from sklearn.preprocessing import StandardScaler
sc1=StandardScaler()
X_train=sc1.fit_transform(X_train)
X_test=sc1.transform(X_test)
sc2=StandardScaler()
y_train=sc2.fit_transform(y_train.reshape(-1,1))
y_test=sc2.transform(y_test.reshape(-1,1))
y_test_inverse=sc2.inverse_transform(y_test)#反归一化

特征选择

filter

name_list=list(data.columns[17:])#是在原列表变量上进行修改，不会返回一个新的修改后的列表
name_list.append('Hardness')
name=pd.Index