MATSim Plans - Sequence to Score¶

Demo of parsing MATSim (experienced) plans for a sequence to score model.

MATSim plans from multiple iterations are processed into plan-scores.

Note that this extracts plan activities and trips. Each sequence component includes duration, model and distance.

In [1]:

from pathlib import Path
import pandas as pd

from pam.read.matsim import load_attributes_map_from_v12, stream_matsim_persons
from pam.utils import datetime_to_matsim_time
from pam.core import Population, Person
from pam.activity import Activity, Plan, Leg

from pathlib import Path import pandas as pd from pam.read.matsim import load_attributes_map_from_v12, stream_matsim_persons from pam.utils import datetime_to_matsim_time from pam.core import Population, Person from pam.activity import Activity, Plan, Leg

In [15]:

dir = Path("data/matsim")

# input paths
output_plans_path = dir / "output_plans.xml"
iter_50 = dir / "output_experienced_plans_ITER50.xml"
assert iter_50.exists()
iter_100 = dir / "output_experienced_plans_ITER100.xml"
assert iter_100.exists()
iter_150 = dir / "output_experienced_plans_ITER150.xml"
assert iter_150.exists()

# output path
output_dir = Path("tmp")
output_dir.mkdir(exist_ok=True)
schedules_path = output_dir / "combined.csv"
attributes_path = output_dir / "attributes.csv"

dir = Path("data/matsim") # input paths output_plans_path = dir / "output_plans.xml" iter_50 = dir / "output_experienced_plans_ITER50.xml" assert iter_50.exists() iter_100 = dir / "output_experienced_plans_ITER100.xml" assert iter_100.exists() iter_150 = dir / "output_experienced_plans_ITER150.xml" assert iter_150.exists() # output path output_dir = Path("tmp") output_dir.mkdir(exist_ok=True) schedules_path = output_dir / "combined.csv" attributes_path = output_dir / "attributes.csv"

In [16]:

streamer50 = stream_matsim_persons(
    iter_50,
    simplify_pt_trips=True,
    # crop=True,
    keep_non_selected=False,
    leg_attributes=True,
    leg_route=True,
)
streamer100 = stream_matsim_persons(
    iter_100,
    simplify_pt_trips=True,
    # crop=True,
    keep_non_selected=False,
    leg_attributes=True,
    leg_route=True,
)
streamer150 = stream_matsim_persons(
    iter_150,
    simplify_pt_trips=True,
    # crop=True,
    keep_non_selected=False,
    leg_attributes=True,
    leg_route=True,
)
streamers = [streamer50, streamer100, streamer150]

streamer50 = stream_matsim_persons( iter_50, simplify_pt_trips=True, # crop=True, keep_non_selected=False, leg_attributes=True, leg_route=True, ) streamer100 = stream_matsim_persons( iter_100, simplify_pt_trips=True, # crop=True, keep_non_selected=False, leg_attributes=True, leg_route=True, ) streamer150 = stream_matsim_persons( iter_150, simplify_pt_trips=True, # crop=True, keep_non_selected=False, leg_attributes=True, leg_route=True, ) streamers = [streamer50, streamer100, streamer150]

In [17]:

def dt_to_min(dt) -> int:
    h, m, s = datetime_to_matsim_time(dt).split(":")
    return (int(h) * 60) + int(m)


def person_to_schedule(person: Person) -> tuple:
    score = person.plan.score
    record = []
    for component in person.plan:
        if isinstance(component, Leg):
            distance = component.euclidean_distance
            mode = component.mode
        else:
            distance = 0
            mode = "NA"
        record.append(
            [
                component.act,
                dt_to_min(component.start_time),
                dt_to_min(component.end_time),
                mode,
                distance,
                score,
            ]
        )
    return person.pid, record


def add_data(record, pid, iteration):
    record = [[pid, iteration] + line for line in record]
    return record

def dt_to_min(dt) -> int: h, m, s = datetime_to_matsim_time(dt).split(":") return (int(h) * 60) + int(m) def person_to_schedule(person: Person) -> tuple: score = person.plan.score record = [] for component in person.plan: if isinstance(component, Leg): distance = component.euclidean_distance mode = component.mode else: distance = 0 mode = "NA" record.append( [ component.act, dt_to_min(component.start_time), dt_to_min(component.end_time), mode, distance, score, ] ) return person.pid, record def add_data(record, pid, iteration): record = [[pid, iteration] + line for line in record] return record

In [18]:

schedules = []
uid = 0
mapper = {}

for iteration, streamer in zip([50, 100, 150], streamers):
    print("iteration: ", iteration)
    for person in streamer:
        pid, record = person_to_schedule(person)
        if record:
            mapper[uid] = pid
            schedules.extend(add_data(record, uid, iteration))
            uid += 1

schedules = pd.DataFrame(
    schedules,
    columns=["pid", "iter", "act", "start", "end", "mode", "distance", "score"],
).set_index("pid")

schedules = [] uid = 0 mapper = {} for iteration, streamer in zip([50, 100, 150], streamers): print("iteration: ", iteration) for person in streamer: pid, record = person_to_schedule(person) if record: mapper[uid] = pid schedules.extend(add_data(record, uid, iteration)) uid += 1 schedules = pd.DataFrame( schedules, columns=["pid", "iter", "act", "start", "end", "mode", "distance", "score"], ).set_index("pid")

iteration:  50
iteration:  100
iteration:  150

In [19]:

attributes = load_attributes_map_from_v12(output_plans_path)
attributes = {k: attributes[v] for k, v in mapper.items()}
attributes = pd.DataFrame(attributes).T
attributes.index.name = "pid"
attributes = attributes.fillna("unknown")
attributes.head()

attributes = load_attributes_map_from_v12(output_plans_path) attributes = {k: attributes[v] for k, v in mapper.items()} attributes = pd.DataFrame(attributes).T attributes.index.name = "pid" attributes = attributes.fillna("unknown") attributes.head()

Out[19]:

	subpopulation	age
pid
0	rich	yes
1	poor	no
2	poor	no
3	poor	no
4	poor	no

In [20]:

schedules.to_csv(schedules_path)
attributes.to_csv(attributes_path)

schedules.to_csv(schedules_path) attributes.to_csv(attributes_path)