from joblib import Parallel, delayed
from sklearn.utils import gen_even_slices
import numpy as np
class TrajectoryData(object):
"""Trajectory data wrapper.
Parameters
----------
attributes : array-like
The names of attributes/features describing trajectory points in the
dataset.
data : array-like, shape: (n_trajectories, n_points, n_features)
The trajectory data.
tids : array-like
The corresponding trajectory IDs of trajectories in ``data``.
labels : array-like (default=None)
The corresponding labels of trajectories in ``data``.
"""
def __init__(self, attributes, data, tids, labels=None):
self.attributes = attributes
self.tids = np.array(tids)
self.labels = np.array(labels) if labels is not None else None
self.data = np.array(data)
self._stats = None
self.tidToIdx = dict(zip(tids, np.r_[0:len(tids)]))
self.labelToIdx = TrajectoryData._get_label_to_idx(labels)
def get_attributes(self):
"""Retrieves the attributes in the dataset.
Returns
-------
attributes : array
An array of length `n_features`.
"""
return self.attributes
def get_tids(self, label=None):
"""Retrieves the trajectory IDs in the dataset.
Parameters
----------
label : int (default=None)
If `None`, then retrieves all trajectory IDs. Otherwise, returns
the IDs corresponding to the given label.
Returns
-------
attributes : array
An array of length `n_trajectories`.
"""
if not label or self.labels is None:
return self.tids
idxs = self.labelToIdx[label]
return self.tids[idxs]
def get_label(self, tid):
"""Retrieves the label for the corresponding tid.
Parameters
----------
tid : int
The trajectory ID.
Returns
-------
label : int or str
The corresponding label.
"""
return self.labels[self.tidToIdx[tid]]
def get_labels(self, unique=False):
"""Retrieves the labels of the trajectories in the dataset.
Parameters
----------
unique : bool (default=False)
If ``True``, then the set of unique labels is returned. Otherwise,
an array with the labels of each individual trajectory is returned.
Returns
-------
labels : array
An array of length `n_trajectories` if `unique=False`, and of
length `n_labels` otherwise.
"""
if unique and self.labels is not None:
return sorted(list(set(self.labels)))
return self.labels
def get_trajectory(self, tid):
"""Retrieves a trajectory from the dataset.
Parameters
----------
tid : int
The trajectory ID.
Returns
-------
trajectory : array, shape: (n_points, n_features)
The corresponding trajectory.
"""
return self.data[self.tidToIdx[tid]]
def get_trajectories(self, label=None):
"""Retrieves multiple trajectories from the dataset.
Parameters
----------
label : int (default=None)
The label of the trajectories to be retrieved. If ``None``, then
all trajectories are retrieved.
Returns
-------
trajectories : array
The trajectories of the given label. If `label=None` or if the
dataset does not contain labels, then all trajectories are
returned.
"""
if not label or self.labels is None:
return self.data
idxs = self.labelToIdx[label]
return self.data[idxs]
def length(self):
"""Returns the number of trajectories in the dataset.
Returns
-------
length : int
Number of trajectories in the dataset.
"""
return len(self.tids)
def merge(self, other, ignore_duplicates=True, inplace=True):
"""Merges this trajectory data with another one. Notice that this
method only works if the datasets have the same set of attributes.
Parameters
----------
other : :class:`trajminer.TrajectoryData`
The dataset to be merged with.
ignore_duplicates : bool (default=True)
If `True`, then trajectory IDs in `other` that already exist in
`self` are ignored. Otherwise, raises an exception when a duplicate
is found.
inplace : bool (default=True)
If `True` modifies the current object, otherwise returns a new
object.
Returns
-------
dataset : :class:`trajminer.TrajectoryData`
The merged dataset. If `inplace=True`, then returns the modified
current object.
"""
if set(self.attributes) != set(other.attributes):
raise Exception("Cannot merge datasets with different sets of " +
"attributes!")
n_attributes = self.attributes
n_tids = self.tids.tolist()
n_labels = self.labels.tolist()
n_data = self.data.tolist()
for tid in other.tids:
if tid in n_tids:
if ignore_duplicates:
continue
raise Exception("tid", tid, "already exists in 'self'!")
n_tids.append(tid)
n_data.append(other.get_trajectory(tid))
if n_labels is not None:
n_labels.append(other.get_label(tid))
if inplace:
self._update(n_attributes, n_data, n_tids, n_labels)
return self
return TrajectoryData(n_attributes, n_data, n_tids, n_labels)
def to_file(self, file, file_type='csv', n_jobs=1):
"""Persists the dataset to a file.
Parameters
----------
file : str
The output file.
file_type : str (default='csv')
The file type. Must be one of `{csv}`.
n_jobs : int (default=1)
The number of parallel jobs.
"""
if file_type == 'csv':
self._to_csv(file, n_jobs)
def stats(self, print_stats=False):
"""Computes statistics for the dataset.
Parameters
----------
print_stats : bool (default=False)
If `True`, stats are printed.
Returns
-------
stats : dict
A dictionary containing the dataset statistics.
"""
if self._stats:
if print_stats:
self._print_stats()
return self._stats
traj_lengths = np.array([len(x) for x in self.data])
points = np.concatenate(self.data)
def count_not_none(arr):
return sum([1 if x is not None else 0 for x in arr])
attr_count = np.array([count_not_none(p) for p in points])
self._stats = {
'attribute': {
'count': len(self.attributes),
'min': attr_count.min(),
'avg': attr_count.mean(),
'std': attr_count.std(),
'max': attr_count.max()
},
'point': {
'count': traj_lengths.sum()
},
'trajectory': {
'count': len(self.data),
'length': {
'min': traj_lengths.min(),
'avg': traj_lengths.mean(),
'std': traj_lengths.std(),
'max': traj_lengths.max()
}
}
}
if self.labels is not None:
unique, counts = np.unique(self.labels, return_counts=True)
self._stats['label'] = {
'count': len(unique),
'min': counts.min(),
'avg': counts.mean(),
'std': counts.std(),
'max': counts.max()
}
if print_stats:
self._print_stats()
return self._stats
def _update(self, attributes, data, tids, labels):
self.tids = np.array(tids)
self.labels = np.array(labels)
self.data = np.array(data)
self.tidToIdx = dict(zip(tids, np.r_[0:len(tids)]))
self.labelToIdx = TrajectoryData._get_label_to_idx(labels)
self._stats = None
def _to_csv(self, file, n_jobs):
lat_lon = -1
tids = self.get_tids()
def build_lines(s):
lines = []
for i in range(s.start, s.stop):
tid = tids[i]
label = self.get_label(tid)
traj = self.get_trajectory(tid)
for p in traj:
if lat_lon > -1:
p[lat_lon] = str(p[lat_lon][0]) + \
',' + str(p[lat_lon][1])
fmt = str(p)[1:-1].replace(', ', ',').replace("'", '')
lines.append(str(tid) + ',' + str(label) + ',' + fmt)
return lines
with open(file, 'w') as out:
header = 'tid,label'
for i, attr in enumerate(self.get_attributes()):
if attr == 'lat_lon':
header += ',lat,lon'
lat_lon = i
else:
header += ',' + attr
out.write(header + '\n')
func = delayed(build_lines)
lines = Parallel(n_jobs=n_jobs, verbose=0)(
func(s) for s in gen_even_slices(len(tids), n_jobs))
lines = np.concatenate(lines)
lines = '\n'.join(lines)
out.write(lines)
out.close()
def _print_stats(self):
print('==========================================================')
print(' STATS ')
print('==========================================================')
print('ATTRIBUTE')
print(' Count: ', self._stats['attribute']['count'])
print(' Min: ', self._stats['attribute']['min'])
print(' Max: ', self._stats['attribute']['max'])
print(' Avg ± Std: %.4f ± %.4f' % (
self._stats['attribute']['avg'], self._stats['attribute']['std']))
print('\nPOINT')
print(' Count: ', self._stats['point']['count'])
print('\nTRAJECTORY')
print(' Count: ', self._stats['trajectory']['count'])
print(' Min length: ',
self._stats['trajectory']['length']['min'])
print(' Max lenght: ',
self._stats['trajectory']['length']['max'])
print(' Avg length ± Std: %.4f ± %.4f' %
(self._stats['trajectory']['length']['avg'],
self._stats['trajectory']['length']['std']))
if self.labels is not None:
print('\nLABEL')
print(' Count: ', self._stats['label']['count'])
print(' Min: ', self._stats['label']['min'])
print(' Max: ', self._stats['label']['max'])
print(' Avg ± Std: %.4f ± %.4f' % (
self._stats['label']['avg'], self._stats['label']['std']))
print('==========================================================')
else:
print('==========================================================')
@staticmethod
def _get_label_to_idx(labels):
labelToIdx = None
if labels is not None:
labelToIdx = {}
for i, label in enumerate(labels):
if label in labelToIdx:
labelToIdx[label].append(i)
else:
labelToIdx[label] = [i]
return labelToIdx
