def encode(self, uid, trajectories, negative_sample=None):
"""standard encoder use the same method as DeepMove
Recode poi id. Encode timestamp with its hour.
Args:
uid ([type]): same as AbstractTrajectoryEncoder
trajectories ([type]): same as AbstractTrajectoryEncoder
trajectory1 = [
(location ID, timestamp, timezone_offset_in_minutes),
(location ID, timestamp, timezone_offset_in_minutes),
.....
]
"""
# 直接对 uid 进行重编码
uid = self.uid
self.uid += 1
encoded_trajectories = []
for index, traj in enumerate(trajectories):
current_loc = []
current_tim = []
current_longi = []
current_lati = []
current_points = []
start_time = parse_time(traj[0][2])
# 以当天凌晨的时间作为计算 time_off 的基准
base_time = cal_basetime(start_time, True)
for point in traj:
loc = point[4]
now_time = parse_time(point[2])
if loc not in self.location2id:
self.location2id[loc] = self.loc_id
self.loc_id += 1
current_points.append(loc)
current_loc.append(self.location2id[loc])
current_lati.append(self.geo_coord[loc][0])
current_longi.append(self.geo_coord[loc][1])
time_code = int(cal_timeoff(now_time, base_time))
if time_code > self.tim_max:
self.tim_max = time_code
current_tim.append(time_code)
# 完成当前轨迹的编码，下面进行输入的形成
trace = []
target = current_loc[-1]
target_tim = current_tim[-1]
current_loc = current_loc[:-1]
current_tim = current_tim[:-1]
lati = self.geo_coord[current_points[-1]][0]
lati = np.array([lati for i in range(len(current_loc))])
longi = self.geo_coord[current_points[-1]][1]
longi = np.array([longi for i in range(len(current_loc))])
current_dis = euclidean_dist(lati - current_lati[:-1],
longi - current_longi[:-1])
trace.append(current_loc)
trace.append(current_tim)
trace.append(target)
trace.append(target_tim)
trace.append(uid)
trace.append(current_dis)
encoded_trajectories.append(trace)
return encoded_trajectories
def _cal_mat1(self, current_tim):
# calculate the temporal relation matrix
mat = np.zeros((self.max_len, self.max_len))
cur_len = len(current_tim)
for i in range(cur_len):
for j in range(cur_len):
off = abs(cal_timeoff(current_tim[i], current_tim[j]))
mat[i][j] = off
if off > self.ex[3]:
self.ex[3] = off
return mat
def cutter_filter(self):
"""
切割后的轨迹存储格式: (dict)
{
uid: [
[
checkin_record,
checkin_record,
...
],
[
checkin_record,
checkin_record,
...
],
...
],
...
}
"""
# load data according to config
traj = pd.read_csv(os.path.join(
self.data_path, '{}.dyna'.format(self.config['dataset'])))
# filter inactive poi
group_location = traj.groupby('location').count()
filter_location = group_location[group_location['time'] >= self.config['min_checkins']]
location_index = filter_location.index.tolist()
traj = traj[traj['location'].isin(location_index)]
user_set = pd.unique(traj['entity_id'])
res = {}
min_session_len = self.config['min_session_len']
max_session_len = self.config['max_session_len']
min_sessions = self.config['min_sessions']
window_size = self.config['window_size']
cut_method = self.config['cut_method']
if cut_method == 'time_interval':
# 按照时间窗口进行切割
for uid in tqdm(user_set, desc="cut and filter trajectory"):
usr_traj = traj[traj['entity_id'] == uid].to_numpy()
sessions = [] # 存放该用户所有的 session
session = [] # 单条轨迹
for index, row in enumerate(usr_traj):
now_time = parse_time(row[2])
if index == 0:
session.append(row.tolist())
prev_time = now_time
else:
time_off = cal_timeoff(now_time, prev_time)
if time_off < window_size and time_off >= 0 and len(session) < max_session_len:
session.append(row.tolist())
else:
if len(session) >= min_session_len:
sessions.append(session)
session = []
session.append(row.tolist())
prev_time = now_time
if len(session) >= min_session_len:
sessions.append(session)
if len(sessions) >= min_sessions:
res[str(uid)] = sessions
elif cut_method == 'same_date':
# 将同一天的 check-in 划为一条轨迹
for uid in tqdm(user_set, desc="cut and filter trajectory"):
usr_traj = traj[traj['entity_id'] == uid].to_numpy()
sessions = [] # 存放该用户所有的 session
session = [] # 单条轨迹
prev_date = None
for index, row in enumerate(usr_traj):
now_time = parse_time(row[2])
now_date = now_time.day
if index == 0:
session.append(row.tolist())
else:
if prev_date == now_date and len(session) < max_session_len:
# 还是同一天
session.append(row.tolist())
else:
if len(session) >= min_session_len:
sessions.append(session)
session = []
session.append(row.tolist())
prev_date = now_date
if len(session) >= min_session_len:
sessions.append(session)
if len(sessions) >= min_sessions:
res[str(uid)] = sessions
else:
# cut by fix window_len used by STAN
if max_session_len != window_size:
raise ValueError('the fixed length window is not equal to max_session_len')
for uid in tqdm(user_set, desc="cut and filter trajectory"):
usr_traj = traj[traj['entity_id'] == uid].to_numpy()
sessions = [] # 存放该用户所有的 session
session = [] # 单条轨迹
for index, row in enumerate(usr_traj):
if len(session) < window_size:
session.append(row.tolist())
else:
sessions.append(session)
session = []
session.append(row.tolist())
if len(session) >= min_session_len:
sessions.append(session)
if len(sessions) >= min_sessions:
res[str(uid)] = sessions
return res
def _load_dyna(self):
"""
轨迹存储格式: (dict)
{
uid: [
[
dyna_record,
dyna_record,
...
],
[
dyna_record,
dyna_record,
...
],
...
],
...
}
"""
# load data according to config
dyna_file = pd.read_csv(os.path.join(
self.data_path, '{}.dyna'.format(self.config['dataset'])))
self._logger.info("Loaded file " + self.config['dataset'] + '.dyna, shape=' + str(dyna_file.shape))
self.dyna_feature_column = {col: i for i, col in enumerate(dyna_file)}
res = dict()
if self.need_cut:
user_set = pd.unique(dyna_file['entity_id'])
min_session_len = self.config['min_session_len']
max_session_len = self.config['max_session_len']
min_sessions = self.config['min_sessions']
window_size = self.config['window_size']
cut_method = self.config['cut_method']
if cut_method == 'time_interval':
# 按照时间窗口进行切割
for uid in tqdm(user_set, desc="cut and filter trajectory"):
usr_traj = dyna_file[dyna_file['entity_id'] == uid]
usr_traj = usr_traj.sort_values(by='time')
usr_traj = usr_traj.reset_index(drop=True)
sessions = [] # 存放该用户所有的 session
traj_id = 0
session = [] # 单条轨迹
for index, row in usr_traj.iterrows():
row['traj_id'] = traj_id
now_time = parse_time(row['time'])
if index == 0:
session.append(row.tolist())
prev_time = now_time
else:
time_off = cal_timeoff(now_time, prev_time)
if time_off < window_size and time_off >= 0 and len(session) < max_session_len:
session.append(row.tolist())
else:
if len(session) >= min_session_len:
sessions.append(session)
traj_id += 1
session = []
session.append(row.tolist())
prev_time = now_time
if len(session) >= min_session_len:
sessions.append(session)
traj_id += 1
if len(sessions) >= min_sessions:
res[str(uid)] = sessions
elif cut_method == 'same_date':
# 将同一天的 check-in 划为一条轨迹
for uid in tqdm(user_set, desc="cut and filter trajectory"):
usr_traj = dyna_file[dyna_file['entity_id'] == uid]
usr_traj = usr_traj.sort_values(by='time')
usr_traj = usr_traj.reset_index(drop=True)
sessions = [] # 存放该用户所有的 session
traj_id = 0
session = [] # 单条轨迹
prev_date = None
for index, row in usr_traj.iterrows():
row['traj_id'] = traj_id
now_time = parse_time(row['time'])
now_date = now_time.day
if index == 0:
session.append(row.tolist().append())
else:
if prev_date == now_date and len(session) < max_session_len:
session.append(row.tolist())
else:
if len(session) >= min_session_len:
sessions.append(session)
traj_id += 1
session = []
session.append(row.tolist())
prev_date = now_date
if len(session) >= min_session_len:
sessions.append(session)
traj_id += 1
if len(sessions) >= min_sessions:
res[str(uid)] = sessions
else:
# cut by fix window_len used by STAN
if max_session_len != window_size:
raise ValueError('the fixed length window is not equal to max_session_len')
for uid in tqdm(user_set, desc="cut and filter trajectory"):
usr_traj = dyna_file[dyna_file['entity_id'] == uid]
usr_traj = usr_traj.sort_values(by='time')
usr_traj = usr_traj.reset_index(drop=True)
sessions = [] # 存放该用户所有的 session
traj_id = 0
session = [] # 单条轨迹
for index, row in usr_traj.iterrows():
row['traj_id'] = traj_id
if len(session) < window_size:
session.append(row.tolist())
else:
sessions.append(session)
traj_id += 1
session = []
session.append(row.tolist())
if len(session) >= min_session_len:
sessions.append(session)
traj_id += 1
if len(sessions) >= min_sessions:
res[str(uid)] = sessions
else:
id_set = set()
for dyna in dyna_file.itertuples():
entity_id = getattr(dyna, "entity_id")
traj_id = getattr(dyna, "traj_id")
if (entity_id, traj_id) in id_set:
continue
id_set.add((entity_id, traj_id))
if entity_id not in res:
res[entity_id] = []
rows = dyna_file[(dyna_file['entity_id'] == entity_id) & (dyna_file['traj_id'] == traj_id)]
rows = rows.sort_values(by='time')
traj = []
for _, row in rows.iterrows():
traj.append(row.tolist())
res[entity_id].append(traj[:])
return res
def cutter_filter(self):
"""
切割后的轨迹存储格式: (dict)
{
uid: [
[
checkin_record,
checkin_record,
...
],
[
checkin_record,
checkin_record,
...
],
...
],
...
}
"""
# load data according to config
traj = pd.read_csv(
os.path.join(self.data_path,
'{}.dyna'.format(self.config['dataset'])))
# filter inactive poi
group_location = traj.groupby('location').count()
filter_location = group_location[
group_location['time'] > self.config['min_checkins']]
location_index = filter_location.index.tolist()
traj = traj[traj['location'].isin(location_index)]
user_set = pd.unique(traj['entity_id'])
res = {}
min_session_len = self.config['min_session_len']
min_sessions = self.config['min_sessions']
window_size = self.config['window_size']
window_type = self.config['window_type']
if window_type == 'time_window':
# 按照时间窗口进行切割
base_zero = window_size > 12
for uid in tqdm(user_set, desc="cut and filter trajectory"):
usr_traj = traj[traj['entity_id'] == uid]
sessions = [] # 存放该用户所有的 session
session = [] # 单条轨迹
# 这里还是使用当地时间吧
start_time = parse_time(usr_traj.iloc[0]['time'])
base_time = cal_basetime(start_time, base_zero)
for index, row in usr_traj.iterrows():
if index == 0:
assert start_time.hour - base_time.hour < window_size
session.append(row.tolist())
else:
now_time = parse_time(row['time'])
time_off = cal_timeoff(now_time, base_time)
if time_off < window_size and time_off >= 0:
session.append(row.tolist())
else:
if len(session) >= min_session_len:
sessions.append(session)
session = []
start_time = now_time
base_time = cal_basetime(start_time, base_zero)
session.append(row.tolist())
if len(session) >= min_session_len:
sessions.append(session)
if len(sessions) >= min_sessions:
# update counter
for s in sessions:
for row in s:
self.counter.update({row[4]: 1})
res[str(uid)] = sessions
else:
# 按照轨迹长度进行划分
for uid in tqdm(user_set, desc="cut and filter trajectory"):
usr_traj = traj[traj['entity_id'] == uid]
sessions = [] # 存放该用户所有的 session
session = [] # 单条轨迹
for index, row in usr_traj.iterrows():
if len(session) < window_size:
session.append(row.tolist())
else:
sessions.append(session)
session = []
session.append(row.tolist())
if len(session) >= min_session_len:
sessions.append(session)
if len(sessions) >= min_sessions:
# update counter
for s in sessions:
for row in s:
self.counter.update({row[4]: 1})
res[str(uid)] = sessions
return res