Usage Examples¶
The examples directory includes both standalone python scripts and jupyter notebooks to help you get started. (NB: Currently there are more scripts than notebooks).
A selection of topics covered:
- Generating transition matrices from data (using various estimators)
- Manipulating transition matrices
- Computing and visualizing credit curves corresponding to a set of transition matrices
- Mapping rating states between different rating systems
Python Scripts¶
The scripts are located in examples/python. For testing purposes all examples can be run using the run_examples.py script located in the root directory. Some scripts have an example flag that selects alternative input data or estimators.
| Script Name | Flag | Input Data | Description |
|---|---|---|---|
| adjust_nr_state.py | 1 | Adjust the NR (not-rated) statistics. | |
| adjust_nr_state.py | 2 | Adjust the NR (not-rated) statistics. | |
| credit_curves.py | Compute and Visualize credit curves | ||
| characterize_datasets.py | Load the available datasets and compute various statistics | ||
| compare_estimators.py | synthetic_data4.csv | Compare the cohort and aalen-johansen estimators on a discrete timestep sample | |
| data_cleaning_example.py | rating_data_raw.csv | Prepare transition data sets (data cleansing) using some provided methods | |
| deterministic_paths.py | Create a transition dataset by replicating give trajectories through a graph | ||
| empirical_transition_matrix.py | 1 | synthetic_data7.csv | Credit Rating Migration example |
| empirical_transition_matrix.py | 2 | synthetic_data8.csv | Simple 2x2 Matrix for testing |
| empirical_transition_matrix.py | 3 | synthetic_data9.csv | Credit Rating Migration example with timestamps in raw date format |
| estimate_matrix.py | rating_data.csv | An end-to-end example of estimating a credit rating matrix from historical data | |
| fix_multiperiod_matrix.py | sp_1981-2016.csv | Detect and solve various pathologies that might be affecting transition matrix data | |
| generate_full_multiperiod_set.py | sp_NR_adjusted.json | Use infinitesimal generator methods to generate a full multi-period matrix set. | |
| generate_synthetic_data.py | 1 | Generate synthetic data. The first set of examples produces duration type data. | |
| generate_synthetic_data.py | 2 | The second set of examples produces cohort type data using markov chain simulation | |
| generate_synthetic_data.py | 3 | The second set of examples produces cohort type data using markov chain simulation | |
| generate_visuals.py | 6 | JLT.json | Plot Transition Probabilities |
| generate_visuals.py | 7 | JLT.json | Logarithmic Sankey Diagram of Credit Migration Rates |
| generate_visuals.py | 5 | scenario_data.csv | Plot Entity Transitions Plot |
| generate_visuals.py | 1 | synthetic_data1.csv | Step Plot of a single observation |
| generate_visuals.py | 4 | synthetic_data3.csv | Entity Transitions Plot |
| generate_visuals.py | 2 | synthetic_data4.csv | Step Plot of individual observations |
| generate_visuals.py | 3 | synthetic_data5.csv | Histogram Plots of transition frequencies |
| matrix_from_cohort_data.py | 3 | synthetic_data4.csv | S&P Style Credit Rating Migration Matrix |
| matrix_from_cohort_data.py | 2 | synthetic_data5.csv | IFRS 9 Style Migration Matrix (Large sample for testing) |
| matrix_from_cohort_data.py | 1 | synthetic_data6.csv | Simplest Absorbing Case for validation |
| matrix_from_duration_data.py | 1 | synthetic_data1.csv | Duration example with limited data (dataset contains only one entity) |
| matrix_from_duration_data.py | 2 | synthetic_data2.csv | Duration example n entities with ~10 observations each, [0,1] state, 50%/50% transition matrix |
| matrix_from_duration_data.py | 3 | synthetic_data3.csv | |
| matrix_lendingclub.py | Estimate a matrix from LendingClub data. Input data are in a special cohort format as the published datasets have some limitations | ||
| matrix_operations.py | Perform various transition matrix operations illustrating the matrix algebra | ||
| matrix_set_lendingclub.py | Estimate a matrix from LendingClub data. Input data are in a special cohort format as the published datasets have some limitations | ||
| matrix_set_operations.py | Perform operations with multi-period transition matrix sequences | ||
| state_space_operations.py | Examples working with state spaces (mappings) |
Jupyter Notebooks¶
- Adjust_NotRated_State.ipynb
- Matrix_Operations.ipynb
- Monthly_from_Annual.ipynb
Open Risk Academy Scripts¶
Additional examples are available in the Open Risk Academy course Analysis of Credit Migration using Python TransitionMatrix. The scripts developed in the course are available here