data_helper.py

import numpy as np
import pandas as pd

def load_timeseries(filename, params):
	"""Load time series dataset"""

	series = pd.read_csv(filename, sep=',', header=0, index_col=0, squeeze=True)
	data = series.values

	adjusted_window = params['window_size']+ 1

	# Split data into windows
	raw = []
	for index in range(len(data) - adjusted_window):
		raw.append(data[index: index + adjusted_window])

	# Normalize data
	result = normalize_windows(raw)

	raw = np.array(raw)
	result = np.array(result)

	# Split the input dataset into train and test
	split_ratio = round(params['train_test_split'] * result.shape[0])
	train = result[:int(split_ratio), :]
	np.random.shuffle(train)

	# x_train and y_train, for training
	x_train = train[:, :-1]
	y_train = train[:, -1]

	# x_test and y_test, for testing
	x_test = result[int(split_ratio):, :-1]
	y_test = result[int(split_ratio):, -1]

	x_train = np.reshape(x_train, (x_train.shape[0], x_train.shape[1], 1))
	x_test = np.reshape(x_test, (x_test.shape[0], x_test.shape[1], 1))  

	x_test_raw = raw[int(split_ratio):, :-1]
	y_test_raw = raw[int(split_ratio):, -1]

	# Last window, for next time stamp prediction
	last_raw = [data[-params['window_size']:]]
	last = normalize_windows(last_raw)
	last = np.array(last)
	last = np.reshape(last, (last.shape[0], last.shape[1], 1))

	return [x_train, y_train, x_test, y_test, x_test_raw, y_test_raw, last_raw, last]

def normalize_windows(window_data):
	"""Normalize data"""

	normalized_data = []
	for window in window_data:
		normalized_window = [((float(p) / float(window[0])) - 1) for p in window]
		normalized_data.append(normalized_window)
	return normalized_data