def check_dist_between_groups(days=None):
if days is None:
days = data.days
out_dir = os.path.join(data.data_out_path, 'group_dist/')
os.makedirs(out_dir, exist_ok=True)
for day in days:
print('-- {} --'.format(day))
t0 = time.clock()
td.initialise(day)
groups = data.read_group_typical_ids(day)
n_groups = len(groups)
dists = np.zeros([n_groups, n_groups])
for i1, i2 in itertools.combinations(range(n_groups), 2):
print('\r{}, {} / {}'.format(i1, i2, n_groups), end='')
dist = td.get_distances(groups.iloc[i1]['id'], groups.iloc[i2]['id'])
dists[i1, i2] = dist
dists[i2, i1] = dist
print('\nDone!')
save_path = os.path.join(out_dir, day)
np.save(save_path, dists)
t1 = time.clock()
print('Take Taken = {}'.format(timedelta(seconds=t1-t0)))
fig, ax = plt.subplots()
ax.pcolormesh(dists)
ax.set_title(day)
def check_dist_within_group():
for day in data.days:
print('-- {} --'.format(day))
td.initialise(day)
groups = data.read_groups(day)
max_max_dist = -1
max_dist_group = -1
dists = []
for group_id, df in groups.groupby('group_id'):
ids = df['id']
max_dist = -1
for id1, id2 in itertools.combinations(ids, 2):
dist = td.get_distances(id1, id2)
dists.append([group_id, dist])
if dist > max_dist:
max_dist = dist
if max_dist > max_max_dist:
max_max_dist = max_dist
max_dist_group = group_id
# print('group {}: max dist = {}'.format(group_id, max_dist))
print('max group dist: group {}, dist = {}'.format(max_dist_group, max_max_dist))
dists = np.array(dists)
fig, ax = plt.subplots()
ax.scatter(dists[:, 0], dists[:, 1], lw=0)
ax.set_title(day)
ax.grid(linestyle='-', alpha=0.2)
ax.yaxis.set_ticks_position('none')
ax.xaxis.set_ticks_position('none')
from pprint import pprint
import itertools
import pandas as pd
import data
import os
import time
from datetime import timedelta
outdir = os.path.join(data.data_out_path, 'groups/')
os.makedirs(outdir, exist_ok=True)
cutoff = 300
for day in data.days:
t0 = time.clock()
print('Day {}'.format(day))
td.initialise(day=day)
# the first id of each group
groups = []
ids = td.all_ids
total = len(ids)
group_df = pd.DataFrame(index=ids, columns=['group_id'])
for i, id in enumerate(ids):
print('\r{}/{}'.format(i, total), end='')
xy = td.df2.loc[:, id]
# as the ids are roughly grouped in the correct groups already,
# we check the first few most recently added groups first,
# which will be at the end
for id2 in itertools.islice(reversed(groups), 30):
# check against the first id
xy2 = td.df2.loc[:, id2]
# calling the private method here because this way we can short circuit and be a little faster
def on_click(event):
if event.inaxes is not ax:
return
cutoff = event.ydata
ax.lines.clear()
ax.plot([0, 1], [cutoff, cutoff], color='red', lw=2)
ids_below = [id for i, id in enumerate(td.all_ids) if dists[i] <= cutoff]
ax3d.clear()
tc.plot_trajectories(ax=ax3d, df=td.df[td.df['id'].isin(ids_below)])
ax3d.set_title('{0} trajectories.\nCutoff = {1:.2f}'.format(
len(ids_below), cutoff))
fig.canvas.draw()
td.initialise(day='Fri')
# initialise display
fig = plt.figure()
ax = fig.add_subplot(1, 2, 1)
ax3d = fig.add_subplot(1, 2, 2, projection='3d')
fig.canvas.mpl_connect('key_press_event', on_key_press)
fig.canvas.mpl_connect('pick_event', on_pick)
fig.canvas.mpl_connect('button_release_event', on_click)
fig.patch.set_color('#EEEEEE')
# plot a random person
plot_random_dists()
plt.show()