Synopsis

TSeqMock is a Python library for temporal sequence simulation.

Features

Multiple Generation Types: Supports three main types of temporal data:
- Event: Sequences of point events in time
- State: Sequences of successive states
- Interval: Sequences of time intervals with variable durations
Configurable Generation: Full customization of statistical distributions, generation methods, and time strategies
Missing Data: Ability to integrate missing values to simulate real-world cases
Multiple Profiles: Creation of multiple data profiles with different configurations

Links

Installation

pip install tseqmock

Usage Examples

Generating Event Sequences

from datetime import timedelta
import numpy as np
from tseqmock.core import TSeqMocker
from tseqmock.generator.base.profile import Profile
from tseqmock.generator.type.event.time_design import EventTimeDesign
from tseqmock.method.base import GenMethod
from tseqmock.time_strategy.base import TimeStrategy
from tseqmock.distribution.base import Distribution

# Initialize the event generator
mock = TSeqMocker("event")

# Configure the generation method
gen_method = GenMethod.init("random")
gen_method.update_settings(vocabulary=np.random.uniform(0, 1, 100))

# Add a profile
mock.add_profile(
    Profile(
        n_seq=10,  # Number of sequences to generate
        sequence_size=4,  # Sizes of the sequences
        data_cols={"event": gen_method},  # Data columns
        missing_data={"event": 0.05},  # Missing data configuration - 5% missing
    )
)

# Configure the time strategy
time_strat = TimeStrategy.init("fixed")
time_strat.update_settings(
    t0_date="2025-01-01",
    # Number of days between sampling (T0, T0+7, T0+7+25, T0+7+25+62)
    # This is equivalent to T0, T0+1 week, T0+1 month, T0+3 months
    sampling_steps=[7, 25, 62],
    granularity="day",
)

# Define time design
mock.set_time_design(
    time_design_settings=EventTimeDesign(
        t0_strategy=time_strat,
        sampling_strategy=time_strat,
    )
)

# Generate data
data = mock()
print(data)
print(mock.summary)

Generating State Sequences

from tseqmock.generator.type.state.time_design import StateTimeDesign

# Initialize the state generator
mock = TSeqMocker("state")

# Configure the generation method
gen_method = GenMethod.init("random")
gen_method.update_settings(vocabulary=["HEALTHY", "SICK", "REMISSION"])

# Add a profile
mock.add_profile(
    Profile(
        n_seq=10,
        sequence_size=[3, 4, 9, 3, 5],
        data_cols={"state": gen_method},
        missing_data={"state": 0},
    )
)

# Configure the time strategy
distrib = Distribution.init("normal", settings={"mu": 3, "sigma": 100})
time_strat = TimeStrategy.init("sequence_specific")
time_strat.update_settings(
    distribution=distrib,
    min_date="1980-01-01",
    max_date="2026-01-01",
)

# Define time design
mock.set_time_design(
    time_design_settings=StateTimeDesign(
        t0_strategy=time_strat,
        sampling_strategy=time_strat,
    )
)

# Generate data
data = mock()
print(data)
print(mock.summary)

Generating Interval Sequences

from tseqmock.generator.type.interval.time_design import IntervalTimeDesign

# Initialize the interval generator
mock = TSeqMocker("interval")

# Configure the generation method
gen_method = GenMethod.init("random")
gen_method.update_settings(vocabulary=[
    "RESTING",
    "RESTAURANT",
    "CITY TOUR",
    "SHOPPING",
])

# Add a profile
mock.add_profile(
    Profile(
        n_seq=10,
        sequence_size=[3, 4, 9, 3, 5],
        data_cols={"interval": gen_method},
    )
)

# Configure the time strategy
time_strat = TimeStrategy.init("sequence_specific")
time_strat.update_settings(
    distribution="uniform",
    min_date="2025-07-01",
    max_date="2025-08-01",
)

# Possible interval durations
interval_durations = [1, 2, 3, 4]
# Define time design
mock.set_time_design(
    time_design_settings=IntervalTimeDesign(
        t0_strategy=time_strat,
        sampling_strategy=time_strat,
        interval_durations=interval_durations,
        granularity="hour",
    )
)

# Generate data
data = mock()
print(data)
print(mock.summary)

Advanced Customization

TSeqMock offers various customization options:

Statistical Distributions: Normal, uniform, etc.
Time Strategies: Sequence-specific, periodic, etc.
Generation Methods: Random, rule-based, etc.

Contributors and Contact

This project was initiated as a student collaboration within the AIstroSight Inria Team.

Student Contributors

The following M1 Bioinformatics students from Université Claude Bernard Lyon 1 contributed to the development during their coursework:

GAGNIEU Thomas
GUEYE Ndeye
RIOU Baptiste
SEBAI Imene

Core Team

Arnaud Duvermy - Lead Developer & Technical Architect
Thomas Guyet - Project Leader & Maintainer

Contact: thomas.guyet@inria.fr