def iet_stats(x, end, start):
"""
Function for obtaining data based on the sequence of events and the IET distribution
Returns:
mu = mean iet with KM
siggma = std iet with KM
burst = burstiness with KM
mu_r = avg relay time with KM
r_frsh = relative freshness
age = age of tie
t_stab = temporal stability of the tie
memory = memory coefficient of the tie
"""
c_norm = 1 #60*60*24
last = (end - start) / c_norm
estimator = events.IntereventTimeEstimator(last, mode='censorall')
w = len(x)
x = [(t - start) / c_norm for t in x]
estimator.add_time_seq(x)
mu = estimator.estimate_moment(1, 'km')
mu_na = estimator.estimate_moment(1, 'naive')
if not mu_na:
mu_na = np.inf
try:
mu2 = estimator.estimate_moment(2, 'km')
sigma = np.sqrt(mu2 - mu**2)
except:
mu2 = np.inf
sigma = np.inf
try:
burst = (sigma - mu) / (sigma + mu)
except:
burst = np.nan
try:
mu_r = (.5) * (mu2 / mu)
except:
mu_r = np.inf
try:
r_frsh = (last - x[-1]) / mu_na
except:
r_frsh = np.inf
age = x[0]
t_stab = x[-1] - x[0]
try:
m = memory_coefficient(x)
except:
m = np.nan
return [w, mu_na, sigma, burst, mu_r, r_frsh, age, t_stab, m]
def _km_versions(x, method, min_date=1167598799, max_date=1177970399.0):
estimator = events.IntereventTimeEstimator(max_date, mode='censorall')
if min_date is not None:
x.insert(0, min_date)
estimator.add_time_seq(x)
mu = estimator.estimate_moment(1, method)
var = estimator.estimate_moment(2, method)
sigma = np.sqrt(var - mu**2)
burst = (sigma - mu) / (sigma + mu)
return round(mu, 2), round(var, 2), round(burst, 4)
def inter_event_times(x, end, start, method='km'):
"""
Obtains inter-event time estimators in terms of days
"""
c_norm = 1.
last = (end - start) / c_norm
estimator = events.IntereventTimeEstimator(last, mode='censorall')
x = [(t - start) / c_norm for t in x]
estimator.add_time_seq(x)
mu = estimator.estimate_moment(1, method)
try:
mu2 = estimator.estimate_moment(2, method)
sigma = np.sqrt(mu2 - mu**2)
except:
mu2 = np.inf
sigma = np.inf
try:
burst = (sigma - mu) / (sigma + mu)
except:
burst = np.nan
try:
avg_relay = (.5) * (mu2 / mu)
except:
avg_relay = np.inf
try:
n = len(x)
r = sigma / mu
burst_c = (np.sqrt(n + 1) * r - np.sqrt(n - 1)) / (
(np.sqrt(n + 1) - 2) * r + np.sqrt(n - 1))
except:
burst_c = np.nan
# RELATIVE FRESHNESS
mu_na = estimator.estimate_moment(1, 'naive')
try:
r_fresh = (last - x[-1]) / mu_na
except:
r_fresh = np.inf
# AGE OF TIE
age_tie = x[0]
# TEMPORAL STABILITY
temp_stab = x[-1] - x[0]
m = memory_coefficient(x)
return [
mu, sigma, burst, burst_c, r_fresh, age_tie, temp_stab, avg_relay, m
]
def mean_inter_event_km(t_seq, mode='km+', max_date=1177970399., moment=1):
"""
Other mode: naive
"""
r = events.IntereventTimeEstimator(max_date, 'censorall')
r.add_time_seq(t_seq)
if mode == 'km+':
mode = 'km'
m = max(r.observed_iets)
r.forward_censored_iets[m] = 1
r.backward_censored_iets[m] = 1
return r.estimate_moment(moment, mode)
def km_residual_intervals(x, method='km', max_date=1177970399.):
estimator = events.IntereventTimeEstimator(max_date, mode='censorall')
if method == 'km+':
method = 'km'
if x[-1] < max_date:
x.append(max_date)
estimator.add_time_seq(x)
m = max(estimator.observed_iets)
estimator.forward_censored_iets[m] = 1
estimator.backward_censored_iets[m] = 1
mu = estimator.estimate_moment(1, method)
return mu
def km_burstiness(x, method='km', max_date=1177970399.):
"""
Burstiness estimator with Mikko's library
method='km' or 'naive', if 'km+', add values to be censored at beginning and end
"""
estimator = events.IntereventTimeEstimator(max_date, mode='censorall')
estimator.add_time_seq(x)
#m = max(estimator.observed_iets)
#estimator.forward_censored_iets[m] = 1
#estimator.backward_censored_iets[m] = 1
mu = estimator.estimate_moment(1, method)
sigma = np.sqrt(estimator.estimate_moment(2, method) - mu**2)
return (sigma - mu) / (sigma + mu)