from sklearn.datasets import load_iris#Get iris data
iris = load_iris() 
from sklearn.model_selection import train_test_split as split #Tool to separate datasets
X_train,X_test,y_train,y_test = split(iris.data, iris.target, train_size = 0.8) #80 of the dataset%For learning, 20%For experimentation
import keras 

#Creating a neural network
 #Dense:Neural network definition class
 #Activateion:Activation function class
from keras.layers import Dense, Activation 
model = keras.models.Sequential()#Make a model Make a container
model.add(Dense(units =32,input_dim = 4 )) #32 intermediate layers, 4 input layers
model.add(Activation('relu'))#Activation function Relu
model.add(Dense(units = 3))#Output layer:Three
model.add(Activation('softmax'))#Activation function softmax

#compile
model.compile( 
loss = 'sparse_categorical_crossentropy', optimizer = 'sgd', metrics = ['accuracy']) 

#Execution of learning
model.fit(X_train,y_train,epochs = 100)#100 times iterative learning

#Perform evaluation
 #Check the correct answer rate of test data
score = model.evaluate(X_test,y_test,batch_size = 1)#loss in score(loss) ,accuracy(accuracy)Vector
#Reference: https://aidiary.hatenablog.com/category/Keras?page=1478696865 
accuracy = score[1] 
print('accuracy="',str(accuracy))#To combine numbers and strings with print, str()Make a string with
#Reference: https://www.javadrive.jp/python/string/index9.html 

 #Check only one data
import numpy as np 
x = np.array([[5.1,3.5,1.4,0.2]])#X_Create an array in the same format as train
r = model.predict(x)#Probability vector
print('Probability per label=',r) 
print('Label with the highest probability=',r.argmax())#argmax()Returns the largest label in the vector

#Output data
json_string = model.to_json() 
#If you want to import a model
#from keras.models import model_fromjson #model = model_from_json/json_string) 
#Save learning parameters
#Install h5py first
model.save_weights('param.hdf5') 
#When reading
#model.load_weight('param.hdf5')  

accuracy=" 0.9666666388511658
Probability per label= [[0.9405338  0.05598015 0.00348606]]
Label with the highest probability= 0

Recommended Posts