from keras.layers import Input, Embedding, LSTM, Dense
from keras.models import Model
import csv
import keras
import pandas as pd
import numpy as np
from keras.constraints import nonneg

notes = pd.read_csv('notes.csv',sep=';',encoding='utf-8')
compare = pd.read_csv('compare.csv',sep=';',encoding='utf-8')
len(compare),len(notes)

meilleur = list(compare.BEST_ETU)

best = pd.DataFrame(columns=['ID_ETU','MATHS','ECO','INFO','ANGLAIS'])
for elt in meilleur:
    best = best.append(notes.iloc[elt:elt+1,:])

moins_bon = []
for i in range(len(compare)):
    if compare.ID_ETU1[i] != compare.BEST_ETU[i]:
        moins_bon.append(compare.ID_ETU1[i])
    else :
        moins_bon.append(compare.ID_ETU2[i])

worst = pd.DataFrame(columns=['ID_ETU','MATHS','ECO','INFO','ANGLAIS'])
for elt in moins_bon:
    worst = worst.append(notes.iloc[elt:elt+1,:])

best = best.drop(columns='ID_ETU')
worst = worst.drop(columns='ID_ETU')

best = best.as_matrix()
worst = worst.as_matrix()

labels = np.array(1).repeat(len(compare))

best.values

labels

input1 = Input(shape=(4,), dtype='float32', name='input1')
input2 = Input(shape=(4,), dtype='float32', name='input2')

dense = Dense(1, activation='linear', W_constraint= nonneg(), bias=False)
dense1 = dense(input1)
dense2 = dense(input2)

merged_vector = keras.layers.concatenate([dense1, dense2], axis=-1)

predictions = Dense(1, activation='sigmoid')(merged_vector)

# We define a trainable model linking the
# tweet inputs to the predictions
model = Model(inputs=[input1,input2], outputs=predictions)

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

model.fit([best, worst], labels, epochs=20)

poids = list(dense.get_weights())

print(poids/np.sum(poids))

premier  = [0.2943,0.1542,0.1197,0.4316]
deuxième = [0.0672,0.5911,0.06136159,0.2802]
troisieme = [0.3024,0.0953,0.5506,0.0515]

# 4928 vs 7791 --> 7791
#moyenne_4928_1 = 11.861*0.2943 + 13.40322*0.1542 + 11.6018*0.1197 + 8.09329*0.4316
moyenne_4928_2 = 11.861*0.0672 + 13.40322*0.5911 + 11.6018*0.0613 + 8.09329*0.2802
moyenne_7791_1 = 12.525*0.2943 + 14.2538*0.1542 + 12.1152*0.1197 + 8.81223*0.4316
moyenne_7791_2 = 12.525*0.0672 + 14.2538*0.5911 + 12.1152*0.0613 + 8.81223*0.2802

print(moyenne_4928_2,moyenne_7791_2)