20191208_神经网络交叉验证

这个单子难点在神经网络交叉验证，之前只会机器学习的交叉验证，借鉴一个微信文章才做出来的，文档链接https://mp.weixin.qq.com/s?__biz=MzA4OTg5NzY3NA==&mid=2649345834&idx=1&sn=3c748d2d3c0ac89395da25a07a75cefa&chksm=880e808fbf7909999e2775254dc6ac0b02fd4fc582977a4de640b6249d838725d6fedf00aead&mpshare=1&scene=23&srcid=1211ZPzaGxnuVALs2XUXGnMx&sharer_sharetime=1576057502530&sharer_shareid=d4f40f7a25def68e84cbad465c76535f#rd

import pandas as pd
import matplotlib.pyplot as plt
%matplotlib inline
import warnings
warnings.filterwarnings('ignore')

test=pd.read_csv('CMP3751M_CMP9772M_ML_Assignment 2-dataset-nuclear_plants_final.csv')
test.head()

	Status	Power_range_sensor_1	Power_range_sensor_2	Power_range_sensor_3	Power_range_sensor_4	Pressure _sensor_1	Pressure _sensor_2	Pressure _sensor_3	Pressure _sensor_4	Vibration_sensor_1	Vibration_sensor_2	Vibration_sensor_3	Vibration_sensor_4
0	Normal	4.5044	0.7443	6.3400	1.9052	29.5315	0.8647	2.2044	6.0480	14.4659	21.6480	15.3429	1.2186
1	Normal	4.4284	0.9073	5.6433	1.6232	27.5032	1.4704	1.9929	5.9856	20.8356	0.0646	14.8813	7.3483
2	Normal	4.5291	1.0199	6.1130	1.0565	26.4271	1.9247	1.9420	6.7162	5.3358	11.0779	25.0914	9.2408
3	Normal	5.1727	1.0007	7.8589	0.2765	25.1576	2.6090	2.9234	6.7485	1.9017	1.8463	28.6640	4.0157
4	Normal	5.2258	0.6125	7.9504	0.1547	24.0765	3.2113	4.4563	5.8411	0.5077	9.3700	34.8122	13.4966

test.describe()

	Power_range_sensor_1	Power_range_sensor_2	Power_range_sensor_3	Power_range_sensor_4	Pressure _sensor_1	Pressure _sensor_2	Pressure _sensor_3	Pressure _sensor_4	Vibration_sensor_1	Vibration_sensor_2	Vibration_sensor_3	Vibration_sensor_4
count	996.000000	996.000000	996.000000	996.000000	996.000000	996.000000	996.000000	996.000000	996.000000	996.000000	996.000000	996.000000
mean	4.999574	6.379273	9.228112	7.355272	14.199127	3.077958	5.749234	4.997002	8.164563	10.001593	15.187982	9.933591
std	2.764856	2.312569	2.532173	4.354778	11.680045	2.126091	2.526136	4.165490	6.173261	7.336233	12.159625	7.282383
min	0.008200	0.040300	2.583966	0.062300	0.024800	0.008262	0.001224	0.005800	0.000000	0.018500	0.064600	0.009200
25%	2.892120	4.931750	7.511400	3.438141	5.014875	1.415800	4.022800	1.581625	3.190292	4.004200	5.508900	3.842675
50%	4.881100	6.470500	9.348000	7.071550	11.716802	2.672400	5.741357	3.859200	6.752900	8.793050	12.185650	8.853050
75%	6.794557	8.104500	11.046800	10.917400	20.280250	4.502500	7.503578	7.599900	11.253300	14.684055	21.835000	14.357400
max	12.129800	11.928400	15.759900	17.235858	67.979400	10.242738	12.647500	16.555620	36.186438	34.867600	53.238400	43.231400

test.isnull().sum()

Status                   0
Power_range_sensor_1     0
Power_range_sensor_2     0
Power_range_sensor_3     0
Power_range_sensor_4     0
Pressure _sensor_1       0
Pressure _sensor_2       0
Pressure _sensor_3       0
Pressure _sensor_4       0
Vibration_sensor_1       0
Vibration_sensor_2       0
Vibration_sensor_3       0
Vibration_sensor_4       0
dtype: int64

def function(a):if 'Normal'in a :return 1else:return 0
test['Status'] = test.apply(lambda x: function(x['Status']), axis = 1)

from sklearn.preprocessing import StandardScaler
x = test.drop(['Status'],axis=1)
y=test['Status']
X_scaler = StandardScaler()
x = X_scaler.fit_transform(x)

print('data shape: {0}; no. positive: {1}; no. negative: {2}'.format(x.shape, y[y==1].shape[0], y[y==0].shape[0]))
from sklearn.model_selection import train_test_split
X_train, X_test, y_train, y_test = train_test_split(x, y, test_size=0.2)

data shape: (996, 12); no. positive: 498; no. negative: 498

import keras
from keras.models import Sequential
from keras.layers import Dense
classifier = Sequential()

Using Theano backend.
WARNING (theano.configdefaults): g++ not available, if using conda: `conda install m2w64-toolchain`
WARNING (theano.configdefaults): g++ not detected ! Theano will be unable to execute optimized C-implementations (for both CPU and GPU) and will default to Python implementations. Performance will be severely degraded. To remove this warning, set Theano flags cxx to an empty string.
WARNING (theano.tensor.blas): Using NumPy C-API based implementation for BLAS functions.

for i in [50,500,1000]:classifier = Sequential()classifier.add(Dense(i, kernel_initializer = 'uniform',activation ='sigmoid', input_dim=12))classifier.add(Dense(1, kernel_initializer = 'uniform',activation ='sigmoid'))classifier.compile(optimizer= 'adam',loss = 'binary_crossentropy',metrics = ['accuracy'])classifier.fit(X_train, y_train, batch_size = 700, epochs = 1)print(' nerve cell{} accuracy rate '.format(i))

Epoch 1/1
796/796 [==============================] - 2s 2ms/step - loss: 0.6935 - accuracy: 0.5113nerve cell50 accuracy rate
Epoch 1/1
796/796 [==============================] - 14s 17ms/step - loss: 0.6934 - accuracy: 0.5113nerve cell500 accuracy rate
Epoch 1/1
796/796 [==============================] - 28s 35ms/step - loss: 0.6924 - accuracy: 0.4975nerve cell1000 accuracy rate

from keras.wrappers.scikit_learn import KerasClassifier
from sklearn.model_selection import cross_val_score

def make_classifier():classifier = Sequential()classiifier.add(Dense(3, kernel_initializer = 'uniform', activation = 'relu', input_dim=12))classiifier.add(Dense(3, kernel_initializer = 'uniform', activation = 'relu'))classifier.add(Dense(1, kernel_initializer = 'uniform', activation = 'sigmoid'))classifier.compile(optimizer= 'adam',loss = 'binary_crossentropy',metrics = ['accuracy'])return classifier

classiifier = KerasClassifier(build_fn = make_classifier,batch_size=100, nb_epoch=300)

accuracies = cross_val_score(estimator = classiifier,X = X_train,y = y_train,cv = 10,n_jobs = -1)

mean = accuracies.mean()

import keras
from keras.wrappers.scikit_learn import KerasClassifier
from sklearn.model_selection import cross_val_score
from keras.models import Sequential
from keras.layers import Dense
for i in [50,500,1000]:classifier = Sequential()classiifier.add(Dense(i,kernel_initializer = 'uniform', activation = 'relu', input_dim=12))classifier.add(Dense(1, kernel_initializer = 'uniform', activation = 'sigmoid'))classifier.compile(optimizer= 'adam',loss = 'binary_crossentropy',metrics = ['accuracy'])classifier.fit(X_train, y_train, batch_size = 100, epochs = 300)print(' nerve cell{} accuracy rate '.format(i))

---------------------------------------------------------------------------NameError                                 Traceback (most recent call last)<ipython-input-14-f1690d4285f8> in <module>()6 for i in [50,500,1000]:7     classifier = Sequential()
----> 8     classiifier.add(Dense(i,kernel_initializer = 'uniform', activation = 'relu', input_dim=12))9     classifier.add(Dense(1, kernel_initializer = 'uniform', activation = 'sigmoid'))10     classifier.compile(optimizer= 'adam',loss = 'binary_crossentropy',metrics = ['accuracy'])NameError: name 'classiifier' is not defined

20191208_神经网络交叉验证相关推荐

基于MATLAB的BR神经网络交叉验证实践
基于MATLAB的BR神经网络交叉验证实践在做项目需要对BR神经网络进行K-Flod交叉验证并最终得到结果,于是写了这个代码. // Author LJS clearload allData.mat ...
Keras训练神经网络进行分类并进行交叉验证（Cross Validation）
Keras训练神经网络进行分类并进行交叉验证(Cross Validation) 交叉验证是在机器学习建立模型和验证模型参数时常用的办法.交叉验证,顾名思义,就是重复的使用数据,把得到的样本数据进行切 ...
神经网络与机器学习笔记—泛化和交叉验证
泛化和交叉验证泛化在反向传播学习中,一般从一个训练样本开始,而且通过网络中装载尽可能多的训练样本来使用反向传播算法计算一个多层感知器的突触权值.希望这样设计的神经网络可以很好的泛化.对于从未在生成 ...
神经网络训练之交叉验证
本文转自 http://blog.csdn.net/acdreamers/article/details/44663891,感谢原作者的付出和分享. 今天来讲一种在机器学习中常用的精度测试方法, ...
【深度学习】(7) 交叉验证、正则化，自定义网络案例：图片分类，附python完整代码
各位同学好,今天和大家分享一下TensorFlow2.0深度学习中的交叉验证法和正则化方法,最后展示一下自定义网络的小案例. 1. 交叉验证交叉验证主要防止模型过于复杂而引起的过拟合,找到使模型泛化 ...
训练数据集如何划分验证测试集？train/test(val/dev) set和交叉验证(cross validation)
普通train/test set 直接将训练数据划分为两部分,一部分用来做训练train set,一部分用来固定作为测试集test set.然后反复更换超参在训练集上进行训练,使用测试集依次测试,进行 ...
遭遇棘手交接_Librosa的城市声音分类-棘手的交叉验证
遭遇棘手交接大纲 (Outline) The goal of this post is two-fold: 这篇文章的目标有两个: I'll show an example of implemen ...
交叉验证和超参数调整：如何优化您的机器学习模型
In the first two parts of this article I obtained and preprocessed Fitbit sleep data, split the data ...
交叉验证python_交叉验证
交叉验证python Cross validation may be any of various model validation techniques that are used to asses ...
深度学习-超参数和交叉验证
一. 1.什么是超参数没接触过机器学习的人可能对这个概念比较模糊.我们可以从两方面来理解 (1)参数值的产生由来超参数是在开始学习过程之前设置值的参数(人为设置),而不是通过训练得到的参数数据. ...

20191208_神经网络交叉验证

20191208_神经网络交叉验证相关推荐

最新文章

热门文章