#!/usr/bin/env python
# coding: utf-8

# # Predicción de Series Temporales NN - Multivariate

# El artículo completo con la explicación detallada en el blog: http://www.aprendemachinelearning.com/

# ## Usaremos Keras y Tensorflow

# Importamos las Librerías que vamos a utilizar

# In[161]:


import pandas as pd
import numpy as np
import matplotlib.pylab as plt
get_ipython().run_line_magic('matplotlib', 'inline')
plt.rcParams['figure.figsize'] = (16, 9)
plt.style.use('fast')

from keras.models import Sequential
from keras.layers import Dense,Activation,Flatten
from sklearn.preprocessing import MinMaxScaler


# ### Cargamos nuestro Dataset

# In[162]:


df = pd.read_csv('time_series.csv',  parse_dates=[0], header=None,index_col=0, names=['fecha','unidades'])
df.head()


# ### Cargamos Datos Categóricos: Día y Mes

# In[163]:


df['weekday']=[x.weekday() for x in df.index]
df['month']=[x.month for x in df.index]
df.head()


# In[164]:


df.describe()


# # Preprocesado de los datos

# In[ ]:


# convert series to supervised learning
def series_to_supervised(data, n_in=1, n_out=1, dropnan=True):
    n_vars = 1 if type(data) is list else data.shape[1]
    df = pd.DataFrame(data)
    cols, names = list(), list()
    # input sequence (t-n, ... t-1)
    for i in range(n_in, 0, -1):
        cols.append(df.shift(i))
        names += [('var%d(t-%d)' % (j+1, i)) for j in range(n_vars)]
    # forecast sequence (t, t+1, ... t+n)
    for i in range(0, n_out):
        cols.append(df.shift(-i))
        if i == 0:
            names += [('var%d(t)' % (j+1)) for j in range(n_vars)]
        else:
            names += [('var%d(t+%d)' % (j+1, i)) for j in range(n_vars)]
    # put it all together
    agg = pd.concat(cols, axis=1)
    agg.columns = names
    # drop rows with NaN values
    if dropnan:
        agg.dropna(inplace=True)
    return agg
 


# In[165]:


PASOS=7

# load dataset
values = df['unidades'].values

# ensure all data is float
values = values.astype('float32')
# normalize features
scaler = MinMaxScaler(feature_range=(-1, 1))
values=values.reshape(-1, 1) # esto lo hacemos porque tenemos 1 sola dimension
scaled = scaler.fit_transform(values)

df['scaled'] = scaled
scaledMerge=df.drop('unidades',axis=1)
#print(scaledMerge.values)

# frame as supervised learning
reframed = series_to_supervised(scaledMerge, PASOS, 1)
reframed.head()


# ## Dividimos en set de Entrenamiento y Validación

# In[166]:


newReframed=reframed.drop(['var1(t)','var2(t)'],axis=1)
print(newReframed.shape)
newReframed.head()


# In[167]:


# split into train and test sets
values = newReframed.values
n_train_days = 315+289 - (30+PASOS)
train = values[:n_train_days, :]
test = values[n_train_days:, :]
# split into input and outputs
x_train, y_train = train[:, :-1], train[:, -1]
x_val, y_val = test[:, :-1], test[:, -1]
# reshape input to be 3D [samples, timesteps, features]
x_train = x_train.reshape((x_train.shape[0], 1, x_train.shape[1]))
x_val = x_val.reshape((x_val.shape[0], 1, x_val.shape[1]))
print(x_train.shape, y_train.shape, x_val.shape, y_val.shape)


# # Creamos el Modelo de Red Neuronal

# ## Utilizaremos una Red Feedforward

# ### Como entradas son 21 columnas (7 pasos por 3 variables)

# In[168]:


def crear_modeloFF():
    model = Sequential() 
    model.add(Dense(PASOS, input_shape=(1,PASOS*3),activation='tanh'))
    model.add(Flatten())
    model.add(Dense(1, activation='tanh'))
    model.compile(loss='mean_absolute_error',optimizer='Adam',metrics=["mse"])
    model.summary()
    return model


# ## Entrenamos nuestra máquina

# In[169]:


EPOCHS=40

model = crear_modeloFF()

history=model.fit(x_train,y_train,epochs=EPOCHS,validation_data=(x_val,y_val),batch_size=PASOS)


# ## Visualizamos Resultados

# In[183]:


results=model.predict(x_val)
print( len(results) )
plt.scatter(range(len(y_val)),y_val,c='g')
plt.scatter(range(len(results)),results,c='r')
plt.title('validate')
plt.show()


# In[186]:


plt.ylim(0.12, 0.35)
plt.plot(history.history['loss'])
plt.title('loss')
plt.plot(history.history['val_loss'])
plt.title('validate loss')
plt.show()


# In[191]:


plt.ylim(0.01, 0.18)
plt.title('Accuracy')
plt.plot(history.history['mean_squared_error'])
plt.show()


# In[173]:


compara = pd.DataFrame(np.array([y_val, [x[0] for x in results]])).transpose()
compara.columns = ['real', 'prediccion']

inverted = scaler.inverse_transform(compara.values)

compara2 = pd.DataFrame(inverted)
compara2.columns = ['real', 'prediccion']
compara2['diferencia'] = compara2['real'] - compara2['prediccion']
compara2.head(10)


# In[174]:


compara2.describe()


# In[192]:


compara2['real'].plot()
compara2['prediccion'].plot()


# # Pronóstico

# A partir de la última semana de noviembre 2018, intentaremos predecir la primer semana de diciembre.

# In[150]:


ultimosDias = df['2018-11-16':'2018-11-30']
ultimosDias


# ## Preparamos los datos para Test

# In[151]:


scaledMerge=ultimosDias.drop('unidades',axis=1)
print(scaledMerge.values)

# frame as supervised learning
reframed = series_to_supervised(scaledMerge, PASOS, 1)
newReframed=reframed.drop(['var1(t)','var2(t)','var3(t)'],axis=1)
newReframed.head(7)


# In[153]:


values = newReframed.values
x_test = values[6:, :]
x_test = x_test.reshape((x_test.shape[0], 1, x_test.shape[1]))
print(x_test.shape)
print(x_test)
ultDiaSemana = newReframed.index[len(newReframed.index)-1].weekday()


# In[154]:


def agregarNuevoValor(x_test,nuevoValor,ultDiaSemana):
    for i in range(x_test.shape[2]-3):
        x_test[0][0][i] = x_test[0][0][i+3]
    ultDiaSemana=ultDiaSemana+1
    if ultDiaSemana>6:
        ultDiaSemana=0
    x_test[0][0][x_test.shape[2]-3]=ultDiaSemana
    x_test[0][0][x_test.shape[2]-2]=12
    x_test[0][0][x_test.shape[2]-1]=nuevoValor
    return x_test,ultDiaSemana


# ## Pronóstico para la "próxima semana"

# In[155]:


results=[]
for i in range(7):
    parcial=model.predict(x_test)
    results.append(parcial[0])
    print('pred',i,x_test)
    x_test,ultDiaSemana=agregarNuevoValor(x_test,parcial[0],ultDiaSemana)
    


# ## Re-Convertimos los resultados

# In[156]:


adimen = [x for x in results]    
print(adimen)
inverted = scaler.inverse_transform(adimen)
inverted


# ## Visualizamos el pronóstico

# In[157]:


prediccion1SemanaDiciembre = pd.DataFrame(inverted)
prediccion1SemanaDiciembre.columns = ['pronostico']
prediccion1SemanaDiciembre.plot()
prediccion1SemanaDiciembre.to_csv('pronostico_multivariate.csv')


# In[158]:


prediccion1SemanaDiciembre


# El artículo completo en www.aprendemachinelearning.com