Plotting retweets over time.

gistfile1.txt

import matplotlib.pyplot as plt
from datetime import datetime, timedelta
import numpy as np
from math import log
import tweepy
import collections
import bisect

DATE_FORMAT = "%Y-%m-%d %H:%M:%S"
INTERVAL = timedelta(hours = 1)
USERNAMES = ["adage", "nytimes", "digg", "wsj", "amazon", "google"]

def normalize(l):
	avg = sum(l) / len(l)
	if avg > 2:
		return log(avg)
	else:
		return avg


auth = tweepy.auth = tweepy.OAuthHandler(consumer_key, consumer_secret)
api = tweepy.API(auth)

dates = []
values = []
for username in USERNAMES:
	test = api.get_user(username)
	ds, vs = zip(*[(t.created_at, t.retweet_count) for t in test.timeline(count = 1000)])
	dates += ds
	values += vs

# Aggregate values with similar datetimes.
beginning_timestamp = min(dates)
interval_grid = [beginning_timestamp + n * INTERVAL for n in range(240)]
bins = collections.defaultdict(list)
for v, d in zip(values, dates):
	index = bisect.bisect(interval_grid, d)
	bins[index].append(v)

# Create a curve of best fit.
x_values = bins.keys()
y_values = [normalize(i) for i in bins.values()]
for day in range(10):
	x_points = np.linspace(0.0, 24.0, 24)
	local_x = [x for x in x_values if x > day * 24 and x < (day + 1) * 24]
	x_indices = [x_values.index(x) for x in local_x]
	local_y = [y_values[y] for y in x_indices]
	local_x = [i - day * 24 for i in local_x]
	if len(local_x) < 3:
		continue
	try:
		coefficients = np.polyfit(local_x, local_y, 2)
		polynomial = np.poly1d(coefficients)
		plt.plot(local_x, local_y, 'x', x_points, polynomial(x_points), '-')
	except:
		continue
plt.show()