time           X1        X2         X3     X4     PREDICT(t+60)
 0 min         63         0          0     1           69
60 min         63         0          2     1          104
120 min         0        10          0     1         pred next value

x_test_n = np.flip(x_test_n, axis=0)
curr_frame = x_test_n[0]
predicted = []
newaxis=[]
for i in range(len(x_test_n)):
   predicted.append(model.predict(curr_frame[newaxis,:])[0,0])
   curr_frame = curr_frame[1:]
   curr_frame = np.insert(curr_frame, [num_time_steps-1], predicted[-1], axis=0)

data= pd.DataFrame(data,columns=['X1','X2','X3','X4'])
data.columns = ['X1', 'X2', 'X3','X4']
pd.options.display.float_format = '{:,.0f}'.format
data = data.dropna ()
y=data['Y'].astype(int)
cols=['X1', 'X2', 'X3','X4']
x=data[cols].astype(int)

data=data.values
scaler_x = preprocessing.MinMaxScaler(feature_range =(0, 1))
x = np.array(x).reshape ((len(x),4 ))
x = scaler_x.fit_transform(x)
scaler_y = preprocessing.MinMaxScaler(feature_range =(0, 1))
y = np.array(y).reshape ((len(y), 1))
y = scaler_y.fit_transform(y)

train_end = 80
x_train=x[0: train_end ,]
x_test=x[train_end +1: ,]
y_train=y[0: train_end]
y_test=y[train_end +1:]
x_train=x_train.reshape(x_train.shape +(1,))
x_test=x_test.reshape(x_test.shape + (1,))

num_time_steps=3
model = Sequential()
model.add(LSTM(4, return_sequences=True,input_shape=(num_time_steps,x_train_n.shape[2])))  # returns a sequence of vectors of dimension 32
model.add(LSTM(8, return_sequences=True))  # returns a sequence of vectors of dimension 32
model.add(LSTM(8))  # return a single vector of dimension 32
model.add(Dense(1))


batchsize = 1
model.compile(loss="mean_squared_error",optimizer="adam")
history = model.fit(x_train,y_train, batch_size = batchsize, nb_epoch=30,validation_data=(x_test, y_test),shuffle =True)

model.reset_states()
x_test_n = np.flip(x_test_n, axis=0)
curr_frame = x_test_n[0]
predicted = []
newaxis=[]
for i in range(len(x_test_n)):
   predicted.append(model.predict(curr_frame[newaxis,:])[0,0])
   curr_frame = curr_frame[1:]
   curr_frame = np.insert(curr_frame, [num_time_steps-1], predicted[-1], axis=0)