CNN2.py

import pandas as pd
from keras.models import Sequential
from keras.layers import *
from gensim.models import Word2Vec
from nltk.corpus import *

#load data
dataset=pd.read_csv('ic.csv', sep=';', engine='python',
    na_values=['NA','?'])
SmilesCanonical=dataset.iloc[:,7:8].values
Observed=dataset.iloc[:,1].values


##Encoding categorical data    : normalement le princip de NLP  text >>>> number
## Encoding the Independent Variable : Independent les variable string dans cette exemple on a SmilesCanonical

print('model de transformation')

modelV = Word2Vec.load("word2vec.model2")  #model2 est bien

SmilesCanonical2 = modelV[modelV.wv.vocab]
#print(Observed[:20])
#print(SmilesCanonical2[:20])

t=modelV.most_similar(positive='Cc1ccccc1N', topn=1)  #pour trouve les molecul similaire
print(t)
print('----------------------------------------------')


input_shape = (356,1)
#premier definie architecture de modele et le nomber de couche

print('model ----------------------------------')
model= Sequential()

model.add(Conv1D (kernel_size = (200), filters = 20, input_shape=input_shape, activation='relu'))
print(model.input_shape)
print(model.output_shape)
model.add(Conv1D(100, 10, activation='relu'))


print(model.output_shape)
model.add(MaxPooling1D(3))

print(model.output_shape)
model.add(Conv1D(160, 10, activation='relu'))
model.add(Conv1D(356, 10, activation='relu'))

print(model.output_shape)
model.add(GlobalAveragePooling1D())

print(model.output_shape)
#la fin de model nta3 chick

model.add(Dense(1))

print(model.output_shape)
#compile le modele



model.compile(loss='mse', optimizer='adam',metrics=['accuracy'])
print('compile model')

#fit model in training data
print('fit model')
model.fit(SmilesCanonical2 ,Observed,batch_size=len(SmilesCanonical2),epochs=100)


#model.fit(data,epochs=10, batch_size=32)