Toy Schedules Generation¶

Generate a population (schedules and attributes). Where every schedule is home -> ACTIVITY -> home, with varying durations. Where ACTIVITY depends on work_status attribute.

In [9]:

from pathlib import Path

import pandas as pd
from numpy.random import choice
import random

from caveat.data.synth import ActivityGen
from caveat.data.utils import generate_population, trace_to_pam
from caveat.evaluate.describe.times import (
    joint_time_distributions_plot,
    times_distributions_plot,
)
from caveat.evaluate.describe.transitions import sequence_prob_plot

from pathlib import Path import pandas as pd from numpy.random import choice import random from caveat.data.synth import ActivityGen from caveat.data.utils import generate_population, trace_to_pam from caveat.evaluate.describe.times import ( joint_time_distributions_plot, times_distributions_plot, ) from caveat.evaluate.describe.transitions import sequence_prob_plot

In [10]:

n = 10000
ps = {
    "employed": [0.8, 0.0, 0.2],
    "student": [0.2, 0.6, 0.2],
    "unemployed": [0.1, 0.1, 0.8],
}

schedules_write_path = Path("tmp/toy_noisy_schedules.csv")
attributes_write_path = Path("tmp/toy_noisy_attributes.csv")

n = 10000 ps = { "employed": [0.8, 0.0, 0.2], "student": [0.2, 0.6, 0.2], "unemployed": [0.1, 0.1, 0.8], } schedules_write_path = Path("tmp/toy_noisy_schedules.csv") attributes_write_path = Path("tmp/toy_noisy_attributes.csv")

In [20]:

pid = []
pids = []
acts = []
starts = []
ends = []
durations = []
work_statuses = []

for i in range(n):
    pid.append(i)
    work_status = random.choice(["employed", "student", "unemployed"])
    probs = ps
    activity = choice(["work", "education", "other"], p=probs[work_status])
    work_statuses.append(work_status)

    budget = 1440
    sequence = ["home"]
    seq_durations = []

    if activity in ["work", "education"]:
        departure_time = random.randint(400, 500)
    else:
        departure_time = random.randint(500, 800)
    seq_durations.append(departure_time)
    budget -= departure_time

    sequence.append(activity)
    if activity == "other":
        act_duration = random.randint(60, 180)
    elif activity == "work":
        act_duration = random.randint(400, 600)
    else:
        act_duration = random.randint(300, 500)
    seq_durations.append(act_duration)
    budget -= act_duration

    if budget > 600:
        sequence.append("home")
        home_duration = random.randint(60, 240)
        seq_durations.append(home_duration)
        budget -= home_duration
        sequence.append("other")
        act_duration = random.randint(60, 180)
        seq_durations.append(act_duration)
        budget -= act_duration

    sequence.append("home")
    seq_durations.append(budget)
    home_duration = budget

    t = 0
    for act, duration in zip(
        sequence,
        seq_durations,
    ):
        pids.append(i)
        acts.append(act)
        starts.append(t)
        t += duration
        ends.append(t)
        durations.append(duration)

schedules = pd.DataFrame(
    {
        "pid": pids,
        "act": acts,
        "start": starts,
        "end": ends,
        "duration": durations,
    }
)

attributes = pd.DataFrame(
    {
        "pid": pid,
        "work_status": work_statuses,
    }
)

print(schedules.head(20))
print(attributes.head())

pid = [] pids = [] acts = [] starts = [] ends = [] durations = [] work_statuses = [] for i in range(n): pid.append(i) work_status = random.choice(["employed", "student", "unemployed"]) probs = ps activity = choice(["work", "education", "other"], p=probs[work_status]) work_statuses.append(work_status) budget = 1440 sequence = ["home"] seq_durations = [] if activity in ["work", "education"]: departure_time = random.randint(400, 500) else: departure_time = random.randint(500, 800) seq_durations.append(departure_time) budget -= departure_time sequence.append(activity) if activity == "other": act_duration = random.randint(60, 180) elif activity == "work": act_duration = random.randint(400, 600) else: act_duration = random.randint(300, 500) seq_durations.append(act_duration) budget -= act_duration if budget > 600: sequence.append("home") home_duration = random.randint(60, 240) seq_durations.append(home_duration) budget -= home_duration sequence.append("other") act_duration = random.randint(60, 180) seq_durations.append(act_duration) budget -= act_duration sequence.append("home") seq_durations.append(budget) home_duration = budget t = 0 for act, duration in zip( sequence, seq_durations, ): pids.append(i) acts.append(act) starts.append(t) t += duration ends.append(t) durations.append(duration) schedules = pd.DataFrame( { "pid": pids, "act": acts, "start": starts, "end": ends, "duration": durations, } ) attributes = pd.DataFrame( { "pid": pid, "work_status": work_statuses, } ) print(schedules.head(20)) print(attributes.head())

    pid        act  start   end  duration
0     0       home      0   669       669
1     0      other    669   803       134
2     0       home    803  1009       206
3     0      other   1009  1101        92
4     0       home   1101  1440       339
5     1       home      0   464       464
6     1  education    464   869       405
7     1       home    869  1440       571
8     2       home      0   581       581
9     2      other    581   645        64
10    2       home    645   786       141
11    2      other    786   942       156
12    2       home    942  1440       498
13    3       home      0   737       737
14    3      other    737   909       172
15    3       home    909  1440       531
16    4       home      0   459       459
17    4       work    459   927       468
18    4       home    927  1440       513
19    5       home      0   460       460
   pid work_status
0    0     student
1    1     student
2    2  unemployed
3    3  unemployed
4    4  unemployed

In [21]:

schedules_write_path.parent.mkdir(exist_ok=True)
attributes_write_path.parent.mkdir(exist_ok=True)
schedules.to_csv(schedules_write_path, index=False)
attributes.to_csv(attributes_write_path, index=False)

schedules_write_path.parent.mkdir(exist_ok=True) attributes_write_path.parent.mkdir(exist_ok=True) schedules.to_csv(schedules_write_path, index=False) attributes.to_csv(attributes_write_path, index=False)

In [22]:

def describe_col(population, col: str) -> pd.DataFrame:
    description = population.groupby("act")[col].describe()[
        ["count", "mean", "std", "min", "max"]
    ]
    description["attribute"] = col
    return description


def describe_cols(population, cols: list[str]) -> pd.DataFrame:
    description = pd.concat(
        [describe_col(population, c) for c in cols], ignore_index=False
    )
    description = description.reset_index().set_index(["attribute", "act"])
    return description


describe_cols(schedules, ["start", "end", "duration"]).round()

def describe_col(population, col: str) -> pd.DataFrame: description = population.groupby("act")[col].describe()[ ["count", "mean", "std", "min", "max"] ] description["attribute"] = col return description def describe_cols(population, cols: list[str]) -> pd.DataFrame: description = pd.concat( [describe_col(population, c) for c in cols], ignore_index=False ) description = description.reset_index().set_index(["attribute", "act"]) return description describe_cols(schedules, ["start", "end", "duration"]).round()

Out[22]:

		count	mean	std	min	max
attribute	act
start	education	2336.0	450.0	29.0	400.0	500.0
	home	24118.0	527.0	454.0	0.0	1242.0
	other	8124.0	773.0	151.0	500.0	1078.0
	work	3658.0	450.0	29.0	400.0	500.0
end	education	2336.0	850.0	64.0	705.0	993.0
	home	24118.0	969.0	424.0	400.0	1440.0
	other	8124.0	893.0	155.0	560.0	1242.0
	work	3658.0	949.0	64.0	801.0	1099.0
duration	education	2336.0	400.0	59.0	300.0	500.0
	home	24118.0	442.0	165.0	60.0	800.0
	other	8124.0	119.0	35.0	60.0	180.0
	work	3658.0	499.0	58.0	400.0	600.0

In [23]:

def time_distributions(population: pd.DataFrame, mapping: dict):
    starts = {k: [] for k in mapping.values()}
    ends = {k: [] for k in mapping.values()}
    durations = {k: [] for k in mapping.values()}
    for act, acts in population.groupby("act"):
        starts[act] = list(acts.start)
        ends[act] = list(acts.end)
        durations[act] = list(acts.duration)
    return starts, ends, durations

def time_distributions(population: pd.DataFrame, mapping: dict): starts = {k: [] for k in mapping.values()} ends = {k: [] for k in mapping.values()} durations = {k: [] for k in mapping.values()} for act, acts in population.groupby("act"): starts[act] = list(acts.start) ends[act] = list(acts.end) durations[act] = list(acts.duration) return starts, ends, durations

In [24]:

_ = times_distributions_plot(schedules, ys={})

_ = times_distributions_plot(schedules, ys={})

In [25]:

_ = joint_time_distributions_plot(schedules, ys={})

_ = joint_time_distributions_plot(schedules, ys={})

In [26]:

_ = sequence_prob_plot(schedules, ys={}, figsize=(8, 6))

_ = sequence_prob_plot(schedules, ys={}, figsize=(8, 6))

In [ ]: