utils.py

import matplotlib.pyplot as plt
from models.lstm import create_time_steps
import numpy as np


def plot_train_history(history, title):
  loss = history.history['loss']
  val_loss = history.history['val_loss']

  epochs = range(len(loss))

  plt.figure()

  plt.plot(epochs, loss, 'b', label='Training loss')
  plt.plot(epochs, val_loss, 'r', label='Validation loss')
  plt.title(title)
  plt.legend()
  plt.grid(True)
  plt.show()


def multi_step_plot(history, true_future, prediction, step):
  plt.figure(figsize=(12, 6))
  num_in = create_time_steps(len(history))
  num_out = len(true_future)

  plt.plot(num_in, np.array(history[:, 0]), label='History')
  plt.plot(np.arange(num_out)/step, np.array(true_future), 'gx',
           label='True Future')
  if prediction.any():
    plt.plot(np.arange(num_out)/step, np.array(prediction), 'rx',
             label='Predicted Future')
  plt.legend(loc='upper left')
  plt.grid(True)
  return plt