def make_transit_periodogram(t, y, dy=0.01):
"""
Plots a periodogram to determine likely period of planet transit candidtaes
in a dataset, based on a box least squared method.
"""
model = BoxLeastSquares(t * u.day, y, dy=0.01)
periodogram = model.autopower(0.2, objective="snr")
plt.figure()
plt.plot(periodogram.period, periodogram.power, 'k')
plt.xlabel('Period [days]')
plt.ylabel('Power')
max_power_i = np.argmax(periodogram.power)
best_fit = periodogram.period[max_power_i]
print('Best Fit Period: {} days'.format(best_fit))
stats = model.compute_stats(periodogram.period[max_power_i],
periodogram.duration[max_power_i],
periodogram.transit_time[max_power_i])
return stats, best_fit
def get_bls_features(x, t):
features = []
# print(x[0])
for i in range(x.shape[0]):
# adapted from http://docs.astropy.org/en/stable/stats/bls.html#peak-statistics
bls = BoxLeastSquares(t, x[i])
periodogram = bls.autopower(
40, minimum_n_transit=5
) # arg is the granularity of considered durations
max_power = np.argmax(periodogram.power)
stats = bls.compute_stats(periodogram.period[max_power],
periodogram.duration[max_power],
periodogram.transit_time[max_power])
# TODO: use dataframe?
features.append([stats[s][0] / stats[s][1] for s in stat_names])
# based on https://arxiv.org/pdf/astro-ph/0206099.pdf
# ratios.append(stats["depth"][0] / stats["depth"][1]) # depth over uncertainty
if (i + 1) % 10 == 0:
print(".", end="")
if (i + 1) % 500 == 0:
print()
print()
return np.array(features)
