gistfile1.py

from datetime import date
import numpy as np
from scipy import polyval, polyfit
from django.utils.datastructures import SortedDict

days, values = np.array([
  (date(2012, 4, 1).toordinal(), 45540),
  (date(2012, 5, 1).toordinal(), 125598),
  (date(2012, 6, 1).toordinal(), 195935),
  (date(2012, 7, 1).toordinal(), 206702),
  (date(2012, 8, 1).toordinal(), 682483-300000),
  (date(2012, 9, 1).toordinal(), 657762),
  (date(2012, 10, 1).toordinal(), 1599805-300000),
]).T

predicted = [
	date(2012, 11, 1).toordinal(),
	date(2012, 12, 1).toordinal(),
	date(2013,  1, 1).toordinal()
]

csv = SortedDict()
for d in days:
	csv[d] = []
for d in predicted:
	csv[d] = []

for degree in [2,3,4,5,6,7]:
	coeff = polyfit(days, values, degree)
	ys = polyval(coeff, days)
	for x,y in zip(days, ys):
		csv[x].append(y)
	for d in predicted:
		csv[d].append(polyval(coeff, d))

for d, vals in csv.items():
	arith_mean = sum(vals) / len(vals)
	harm_mean = len(vals) / sum(map(lambda y: 1/y, vals))
	geom_mean = pow(reduce(lambda x,y: x*y, vals), 1/float(len(vals)))
	print '%s,%s,%d,%d,%d' % (date.fromordinal(d), ','.join(map(str,map(int,vals))), arith_mean, harm_mean, geom_mean)