Python Forum

for i in range(24):


  LSTM_matrix.append(y_pred)

 #build a 0 matrix to populate with LSTM output
def build_matrix(rows, cols):
    matrix = []

    for r in range(0, rows):
        matrix.append([0 for c in range(0, cols)])

    return matrix

LSTM_matrix = build_matrix(31, 24)

#i need to iterate over the dataframe rows one step at a time
for i in range(24):
    #24 is based off 0 indexing and looping up until dec-2017
    df2 = df.iloc[:88+i]
    number_column1 = df2.loc[:,'NZ']
    number_column2 = df2.loc[:,'Unemployment']
    number_column3 = df2.loc[:,'CPI']
    number_column4 = df2.loc[:,'Export prices']
    number_column5 = df2.loc[:,'Import prices']

    numbers1 = number_column1.values
    numbers2 = number_column2.values
    numbers3 = number_column3.values
    numbers4 = number_column4.values
    numbers5 = number_column5.values

    # convert to [rows, columns] structure
    numbers1 = numbers1.reshape((len(numbers1), 1))
    numbers2 = numbers2.reshape((len(numbers2), 1))
    numbers3 = numbers3.reshape((len(numbers3), 1))
    numbers4 = numbers4.reshape((len(numbers4), 1))
    numbers5 = numbers5.reshape((len(numbers5), 1))
    
    # horizontally stack columns
    dataset = hstack((numbers2, numbers3,numbers4,numbers5, numbers1))

    # covert into input/output
    X, y = split_sequences(dataset, n_steps_in, n_steps_out)
    # the dataset knows the number of features, e.g. 2
    n_features = X.shape[2]
    
    # define model
    model = Sequential()
    model.add(LSTM(120, activation='relu', return_sequences=True, input_shape=(n_steps_in, n_features)))
    model.add(LSTM(120, activation='relu'))
    model.add(Dense(n_steps_out))
    model.compile(optimizer='adam', loss='mse')
    
    # fit model
    history = model.fit(X, y, epochs=200, verbose=0)
    
    x_input = df.iloc[80+i:88+i]
    test_column2 = x_input.loc[:,'Unemployment']
    test_column3 = x_input.loc[:,'CPI']
    test_column4 = x_input.loc[:,'Export prices']
    test_column5 = x_input.loc[:,'Import prices']



    test2 = test_column2.values
    test3 = test_column3.values
    test4 = test_column4.values
    test5 = test_column5.values


    # convert to [rows, columns] structure
    test2 = test2.reshape((len(test2), 1))
    test3 = test3.reshape((len(test3), 1))
    test4 = test4.reshape((len(test4), 1))
    test5 = test5.reshape((len(test5), 1))


    # horizontally stack columns
    x_input = hstack((test2, test3,test4,test5))
    
    #make prediction
    x_input = x_input.reshape((1, n_steps_in, n_features))
    y_pred = model.predict(x_input, verbose=0)

    
    y_pred = y_pred[0:,8:16]
    y_pred = y_pred.transpose()
    
    LSTM_matrix.append(y_pred)
    print(y_pred)