test.py

from pandas import Series
from pandas import read_csv
from statsmodels.tsa.arima_model import ARIMA
import numpy

# create a differenced series
def difference(dataset, interval=1):
	diff = list()
	for i in range(interval, len(dataset)):
		value = dataset[i] - dataset[i - interval]
		diff.append(value)
	return numpy.array(diff)

# invert differenced value
def inverse_difference(history, yhat, interval=1):
	return yhat + history[-interval]

# load dataset
series = Series.from_csv('dataset.csv', header=None)
# seasonal difference
X = series.values
days_in_year = 365
differenced = difference(X, days_in_year)

# fit model
model = ARIMA(differenced, order=(7,0,1))
model_fit = model.fit(disp=0)
# multi-step out-of-sample forecast
forecast = model_fit.forecast(steps=7)[0]
# invert the differenced forecast to something usable
history = [x for x in X]
day = 1
i = 0
validation = read_csv('validation.csv', header=None)

for yhat in forecast:
	while i in range(len(validation)):
		inverted = inverse_difference(history, yhat, days_in_year)
		print('Day %d Predicted: %s      Actual: %s' % (day, inverted, validation[1][i]))
		history.append(inverted)
		day += 1
		i += 1