7. Latent class model

Bayesian estimation of a discrete mixture of logit (or latent class model).

Michel Bierlaire, EPFL Mon Nov 03 2025, 13:36:51

from pathlib import Path

from IPython.core.display_functions import display

See the data processing script: Data preparation for Swissmetro.

from swissmetro_data import (
    CAR_AV_SP,
    CAR_CO_SCALED,
    CAR_TT_SCALED,
    CHOICE,
    SM_AV,
    SM_COST_SCALED,
    SM_TT_SCALED,
    TRAIN_AV_SP,
    TRAIN_COST_SCALED,
    TRAIN_TT_SCALED,
    database,
)

from biogeme.bayesian_estimation import (
    BayesianResults,
    BayesianResultsSummary,
    get_pandas_estimated_parameters,
)
from biogeme.biogeme import BIOGEME
from biogeme.expressions import Beta, log
from biogeme.models import logit

Parameters to be estimated.

asc_car = Beta('asc_car', 0, None, None, 0)
asc_train = Beta('asc_train', 0, None, None, 0)
asc_sm = Beta('asc_sm', 0, None, None, 1)
b_time = Beta('b_time', 0, None, None, 0)
b_cost = Beta('b_cost', 0, None, None, 0)

Class membership probability.

prob_class1 = Beta('prob_class1', 0.5, 0, 1, 0)
prob_class2 = 1 - prob_class1

Definition of the utility functions for latent_old class 1, where the time coefficient is zero.

v_train_class_1 = asc_train + b_cost * TRAIN_COST_SCALED
v_swissmetro_class_1 = asc_sm + b_cost * SM_COST_SCALED
v_car_class_1 = asc_car + b_cost * CAR_CO_SCALED

Associate utility functions with the numbering of alternatives.

v_class_1 = {1: v_train_class_1, 2: v_swissmetro_class_1, 3: v_car_class_1}

Definition of the utility functions for latent_old class 2, whete the time coefficient is estimated.

v_train_class_2 = asc_train + b_time * TRAIN_TT_SCALED + b_cost * TRAIN_COST_SCALED
v_swissmetro_class_2 = asc_sm + b_time * SM_TT_SCALED + b_cost * SM_COST_SCALED
v_car_class_2 = asc_car + b_time * CAR_TT_SCALED + b_cost * CAR_CO_SCALED

Associate utility functions with the numbering of alternatives.

v_class_2 = {1: v_train_class_2, 2: v_swissmetro_class_2, 3: v_car_class_2}

Associate the availability conditions with the alternatives.

av = {1: TRAIN_AV_SP, 2: SM_AV, 3: CAR_AV_SP}

The choice model is a discrete mixture of logit, with availability conditions

choice_probability_class_1 = logit(v_class_1, av, CHOICE)
choice_probability_class_2 = logit(v_class_2, av, CHOICE)
prob = (
    prob_class1 * choice_probability_class_1 + prob_class2 * choice_probability_class_2
)
log_probability = log(prob)

Create the Biogeme object

the_biogeme = BIOGEME(database, log_probability)
the_biogeme.model_name = 'b07_discrete_mixture'

Estimate the posterior distribution of the parameters, or read the results if already available.

yaml_file = Path('saved_results') / f'{the_biogeme.model_name}.yaml'
try:
    summary_results = BayesianResultsSummary.from_yaml_file(filename=yaml_file)
except FileNotFoundError:
    results: BayesianResults = the_biogeme.bayesian_estimation()
    summary_results = results.to_summary()
print(summary_results.short_summary())
Sample size                                              6768
Sampler                                                  NUTS
Number of chains                                         4
Number of draws per chain                                2000
Total number of draws                                    8000
Acceptance rate target                                   0.9
Run time                                                 0:01:17.790031
Posterior predictive log-likelihood (sum of log mean p)  -5208.05
Expected log-likelihood E[log L(Y|θ)]                    -5210.99
Best-draw log-likelihood (posterior upper bound)         -5208.53
LOO (Leave-One-Out Cross-Validation)                     -5213.94
LOO Standard Error                                       53.21
Effective number of parameters (p_LOO)                   5.89

Present the parameter estimates in a pandas table.

pandas_results = get_pandas_estimated_parameters(
    estimation_results=summary_results,
)
display(pandas_results)
          Name  Value (mean)  ...   ESS (bulk)   ESS (tail)
0    asc_train     -0.399529  ...  4062.552923  5060.974518
1       b_cost     -1.266278  ...  5583.692324  5172.725233
2      asc_car      0.123778  ...  4449.233525  4867.994695
3       b_time     -2.813642  ...  3765.422393  4294.823690
4  prob_class1      0.252521  ...  4497.808463  4639.971249

[5 rows x 12 columns]

Report the variables stored in the Bayesian estimation results.

display(summary_results.report_stored_variables())
             group           variable                dims            shape
0    constant_data          CAR_AV_SP               [obs]           [6768]
1    constant_data      CAR_CO_SCALED               [obs]           [6768]
2    constant_data      CAR_TT_SCALED               [obs]           [6768]
3    constant_data             CHOICE               [obs]           [6768]
4    constant_data              SM_AV               [obs]           [6768]
5    constant_data     SM_COST_SCALED               [obs]           [6768]
6    constant_data       SM_TT_SCALED               [obs]           [6768]
7    constant_data        TRAIN_AV_SP               [obs]           [6768]
8    constant_data  TRAIN_COST_SCALED               [obs]           [6768]
9    constant_data    TRAIN_TT_SCALED               [obs]           [6768]
10  log_likelihood            _choice  [chain, draw, obs]  [4, 2000, 6768]
11       posterior            asc_car       [chain, draw]        [4, 2000]
12       posterior          asc_train       [chain, draw]        [4, 2000]
13       posterior             b_cost       [chain, draw]        [4, 2000]
14       posterior             b_time       [chain, draw]        [4, 2000]
15       posterior           log_like  [chain, draw, obs]  [4, 2000, 6768]
16       posterior        prob_class1       [chain, draw]        [4, 2000]
17           prior            asc_car       [chain, draw]        [1, 2000]
18           prior          asc_train       [chain, draw]        [1, 2000]
19           prior             b_cost       [chain, draw]        [1, 2000]
20           prior             b_time       [chain, draw]        [1, 2000]
21           prior           log_like  [chain, draw, obs]  [1, 2000, 6768]
22           prior        prob_class1       [chain, draw]        [1, 2000]
23    sample_stats    acceptance_rate       [chain, draw]        [4, 2000]
24    sample_stats          diverging       [chain, draw]        [4, 2000]
25    sample_stats             energy       [chain, draw]        [4, 2000]
26    sample_stats                 lp       [chain, draw]        [4, 2000]
27    sample_stats            n_steps       [chain, draw]        [4, 2000]
28    sample_stats          step_size       [chain, draw]        [4, 2000]
29    sample_stats         tree_depth       [chain, draw]        [4, 2000]

Total running time of the script: (0 minutes 1.134 seconds)

Gallery generated by Sphinx-Gallery