from joblib import Parallel, delayed
from sklearn.utils import gen_even_slices
import numpy as np
from ..trajectory_data import TrajectoryData
class TrajectorySegmenter(object):
"""Trajectory segmenter.
Parameters
----------
attributes : array-like
The attributes of a trajectory dataset.
thresholds : dict (default=None)
A dictionary with callables for the attributes that will be used in
the segmentation (e.g. dict['time'] is the callable for attribute
`time`). A callable takes as input two attribute values and outputs
`True` if the trajectory should be segmented and `False` otherwise.
For instance, suppose we have a dataset with a `time` attribute which
represents the minutes from midnight when a trajectory point was
recorded. If we'd like to segment trajectories every time there is a
distance of more than 60 minutes between points, the callable for time
would be defined as::
lambda x, y: abs(y - x) > 60 if y >= x else 60 * 24 - x + y > 60
If ``None``, then trajectories are segmented whenever two attribute
values are different (this behavior changes according to the `mode`
parameter).
mode : str (default='strict')
A string in {'strict', 'any'}:
- If 'strict', then trajectories are segmented when thresholds are
`True` for all attributes.
- If 'any', then trajectories are segmented when at least one
threshold is `True` for any attribute.
ignore_missing : bool (default=False)
If `False`, then trajectories are segmented whenever a missing value is
found (this behavior changes according to the `mode` parameter).
n_jobs : int (default=1)
The number of parallel jobs.
"""
def __init__(self, attributes, thresholds=None, mode='strict',
ignore_missing=False, n_jobs=1):
self.attributes = attributes
self.thresholds = thresholds
self.mode = mode
self.n_jobs = n_jobs
self.ignore_missing = ignore_missing
if not thresholds:
self.thresholds = {}
for attr in attributes:
self.thresholds[attr] = \
lambda x, y: not self.ignore_missing if not x or not y \
else x != y
def fit_transform(self, X):
"""Fit and segment trajectories.
Parameters
----------
X : :class:`trajminer.TrajectoryData`
Input dataset to segment.
Returns
-------
X_out : :class:`trajminer.TrajectoryData`
Segmented dataset.
"""
tids = X.get_tids()
def segment(X, s):
def check_segment(p1, p2):
b = []
for i, attr in enumerate(self.attributes):
f = self.thresholds[attr]
b.append(f(p1[i], p2[i]))
return np.any(b) if self.mode == 'any' else np.all(b)
ret = []
for t in range(s.start, s.stop):
subret = []
traj = X.get_trajectory(tids[t])
s = [traj[0]]
for i in range(1, len(traj)):
if check_segment(traj[i - 1], traj[i]):
subret.append(s)
s = [traj[i]]
else:
s.append(traj[i])
subret.append(s)
ret.append(subret)
return ret
func = delayed(segment)
segments = Parallel(n_jobs=self.n_jobs, verbose=0)(
func(X, s) for s in gen_even_slices(len(X.get_trajectories()),
self.n_jobs))
labels = X.get_labels()
segments = np.squeeze(segments)
new_labels = None
if labels is not None:
new_labels = []
for idx, l in enumerate(labels):
new_labels.extend(np.full(len(segments[idx]), l))
segments = np.squeeze(segments)
new_tids = np.r_[1:len(segments) + 1]
return TrajectoryData(attributes=X.get_attributes(),
data=segments,
tids=new_tids,
labels=new_labels)
