Generators SubPackage

This subpackage implements test data generation

transitionMatrix.generators contents

transitionMatrix.generators.dataset_generators.deterministic(sequences, replication_count)

Generate a transition dataset from a given sequence. Each element in the list is a different transition profile through the state space starting with the initial observation. Replicating the sequences a number of times generates a statistical sample (each replication being a different entity). This allows controlled testing.

Parameters:

• sequences – a list of sequences to replicate

• replication_count – the number of replications

Returns:

transitionMatrix.generators.dataset_generators.exponential_transitions(statespace, n, sample, rate, data_format='Compact')

Generate continuous time events from exponential distribution and uniform sampling from state space. Suitable for testing cohorting algorithms and duration based estimators.

The data are sorted by entity ID, then by time of occurrence T. The first entry per entity indicates the state up to that timepoint. The format is a sequence of triples (ID, Time, State)

Parameters:

• statespace – The state space to use for the simulation

• n (int) – The number of distinct entities to simulate

• sample (int) – The number of events to simulate

• rate (float) – The event rate

Returns:

transition events

Return type:

pandas dataframe

Note

May generate successive events in the same state

transitionMatrix.generators.dataset_generators.long_format(statespace, mymatrix, n, timesteps, mode='Canonical')

Generate continuous events from a markov chain matrix in long data format. Suitable for testing duration based estimators

Parameters:

• statespace (object) – The state space to use for the simulation

• mymatrix (object) – The transition matrix to use for the simulation

• n (int) – The number of distinct entities to simulate

• timesteps (int) – The number of timesteps to simulate (including the initial state)

• mode (str) – How to encode the data (Canonical -> From/To, Compact -> State)

Returns:

generated dataset

Return type:

pandas dataframe

transitionMatrix.generators.dataset_generators.markov_chain(statespace, transitionmatrix, n, timesteps)

Generate discrete events from a markov chain matrix in Compact data format. Suitable for testing cohort based estimators (each time step is a cohort)

Parameters:

• n (int) – The number of distinct entities to simulate

• timesteps (int) – The number of timesteps to simulate (including initial state)

Returns:

the message id

Return type:

pandas dataframe

transitionMatrix.generators.dataset_generators.portfolio_labels(statespace, n)

Generate a collection of credit rating states emulating a snapshot of portfolio data. Suitable for mappings and transformations of credit rating states

Parameters:

n (int) – The number of distinct entities to generate

Returns:

the collection

Return type:

list