diff --git a/choice_learn/models/latent_class_base_model.py b/choice_learn/models/latent_class_base_model.py
index 1b70f949..adfd7550 100644
--- a/choice_learn/models/latent_class_base_model.py
+++ b/choice_learn/models/latent_class_base_model.py
@@ -3,6 +3,7 @@
import time
import numpy as np
+import pandas as pd
import tensorflow as tf
import tqdm
@@ -930,3 +931,37 @@ def get_latent_classes_weights(self):
Latent classes weights/probabilities
"""
return tf.nn.softmax(tf.concat([[tf.constant(0.0)], self.latent_logits], axis=0))
+
+ def compute_report(self, choice_dataset):
+ """Compute a report of the estimated weights.
+
+ Parameters
+ ----------
+ choice_dataset : ChoiceDataset
+ ChoiceDataset used for the estimation of the weights that will be
+ used to compute the Std Err of this estimation.
+
+ Returns
+ -------
+ pandas.DataFrame
+ A DF with estimation, Std Err, z_value and p_value for each coefficient.
+ """
+ reports = []
+ for i, model in enumerate(self.models):
+ compute = getattr(model, "compute_report", None)
+ if callable(compute):
+ report = model.compute_report(choice_dataset)
+ report["Latent Class"] = i
+ reports.append(report)
+ else:
+ raise ValueError(f"{i}-th model {model} does not have a compute_report method.")
+ return pd.concat(reports, axis=0, ignore_index=True)[
+ [
+ "Latent Class",
+ "Coefficient Name",
+ "Coefficient Estimation",
+ "Std. Err",
+ "z_value",
+ "P(.>z)",
+ ]
+ ]
diff --git a/notebooks/models/latent_class_model.ipynb b/notebooks/models/latent_class_model.ipynb
index 62ad7d60..82b4ed8f 100644
--- a/notebooks/models/latent_class_model.ipynb
+++ b/notebooks/models/latent_class_model.ipynb
@@ -39,6 +39,7 @@
"\n",
"os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"\"\n",
"\n",
+ "import matplotlib as mpl\n",
"import numpy as np\n",
"import pandas as pd\n",
"\n",
@@ -106,9 +107,254 @@
"print(f\"Negative Log-Likelihood: {nll}\")"
]
},
+ {
+ "cell_type": "code",
+ "execution_count": null,
+ "metadata": {
+ "keep_output": true
+ },
+ "outputs": [
+ {
+ "name": "stdout",
+ "output_type": "stream",
+ "text": [
+ "Using L-BFGS optimizer, setting up .fit() function\n",
+ "Using L-BFGS optimizer, setting up .fit() function\n",
+ "Using L-BFGS optimizer, setting up .fit() function\n"
+ ]
+ },
+ {
+ "name": "stderr",
+ "output_type": "stream",
+ "text": [
+ "/var/folders/zz/r1py7zhj35q75v09h8_42nzh0000gp/T/ipykernel_67121/1263996749.py:4: MatplotlibDeprecationWarning: The get_cmap function was deprecated in Matplotlib 3.7 and will be removed two minor releases later. Use ``matplotlib.colormaps[name]`` or ``matplotlib.colormaps.get_cmap(obj)`` instead.\n",
+ " cmap = mpl.cm.get_cmap(\"Set1\")\n"
+ ]
+ },
+ {
+ "data": {
+ "text/html": [
+ "\n",
+ "\n",
+ " \n",
+ " \n",
+ " | | \n",
+ " Latent Class | \n",
+ " Coefficient Name | \n",
+ " Coefficient Estimation | \n",
+ " Std. Err | \n",
+ " z_value | \n",
+ " P(.>z) | \n",
+ "
\n",
+ " \n",
+ " \n",
+ " \n",
+ " | 0 | \n",
+ " 0 | \n",
+ " Weights_items_features_0 | \n",
+ " -0.675645 | \n",
+ " 0.023987 | \n",
+ " -28.167109 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 1 | \n",
+ " 0 | \n",
+ " Weights_items_features_1 | \n",
+ " -0.060604 | \n",
+ " 0.008162 | \n",
+ " -7.424849 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 2 | \n",
+ " 0 | \n",
+ " Weights_items_features_2 | \n",
+ " 1.851951 | \n",
+ " 0.054914 | \n",
+ " 33.724579 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 3 | \n",
+ " 0 | \n",
+ " Weights_items_features_3 | \n",
+ " 1.322549 | \n",
+ " 0.048159 | \n",
+ " 27.462420 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 4 | \n",
+ " 0 | \n",
+ " Weights_items_features_4 | \n",
+ " -5.857089 | \n",
+ " 0.191162 | \n",
+ " -30.639460 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 5 | \n",
+ " 0 | \n",
+ " Weights_items_features_5 | \n",
+ " -6.513206 | \n",
+ " 0.195680 | \n",
+ " -33.285046 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 6 | \n",
+ " 1 | \n",
+ " Weights_items_features_0 | \n",
+ " -1.817566 | \n",
+ " 0.077771 | \n",
+ " -23.370796 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 7 | \n",
+ " 1 | \n",
+ " Weights_items_features_1 | \n",
+ " -1.726365 | \n",
+ " 0.058838 | \n",
+ " -29.340986 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 8 | \n",
+ " 1 | \n",
+ " Weights_items_features_2 | \n",
+ " 3.696567 | \n",
+ " 0.160258 | \n",
+ " 23.066404 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 9 | \n",
+ " 1 | \n",
+ " Weights_items_features_3 | \n",
+ " 4.111840 | \n",
+ " 0.157179 | \n",
+ " 26.160225 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 10 | \n",
+ " 1 | \n",
+ " Weights_items_features_4 | \n",
+ " -26.693516 | \n",
+ " 3.274723 | \n",
+ " -8.151381 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 11 | \n",
+ " 1 | \n",
+ " Weights_items_features_5 | \n",
+ " -14.925840 | \n",
+ " 0.634699 | \n",
+ " -23.516403 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 12 | \n",
+ " 2 | \n",
+ " Weights_items_features_0 | \n",
+ " -2.104791 | \n",
+ " 0.104296 | \n",
+ " -20.181009 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 13 | \n",
+ " 2 | \n",
+ " Weights_items_features_1 | \n",
+ " -1.652622 | \n",
+ " 0.073820 | \n",
+ " -22.387188 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 14 | \n",
+ " 2 | \n",
+ " Weights_items_features_2 | \n",
+ " -5.554287 | \n",
+ " 0.245318 | \n",
+ " -22.641151 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 15 | \n",
+ " 2 | \n",
+ " Weights_items_features_3 | \n",
+ " -13.565555 | \n",
+ " 0.544168 | \n",
+ " -24.928965 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 16 | \n",
+ " 2 | \n",
+ " Weights_items_features_4 | \n",
+ " -9.794930 | \n",
+ " 0.631004 | \n",
+ " -15.522781 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ " | 17 | \n",
+ " 2 | \n",
+ " Weights_items_features_5 | \n",
+ " -12.126673 | \n",
+ " 0.681118 | \n",
+ " -17.804060 | \n",
+ " 0.000000 | \n",
+ "
\n",
+ " \n",
+ "
\n"
+ ],
+ "text/plain": [
+ ""
+ ]
+ },
+ "execution_count": null,
+ "metadata": {},
+ "output_type": "execute_result"
+ }
+ ],
+ "source": [
+ "report = lc_model.compute_report(elec_dataset)\n",
+ "\n",
+ "def format_color_groups(df):\n",
+ " cmap = mpl.cm.get_cmap(\"Set1\")\n",
+ " colors = [mpl.colors.rgb2hex(cmap(i)) for i in range(cmap.N)]\n",
+ " x = df.copy()\n",
+ " factors = list(x['Latent Class'].unique())\n",
+ " i = 0\n",
+ " for factor in factors:\n",
+ " style = f'background-color: {colors[i]}'\n",
+ " x.loc[x['Latent Class'] == factor, :] = style\n",
+ " i += 1\n",
+ " return x\n",
+ "\n",
+ "report.style.apply(format_color_groups, axis=None)"
+ ]
+ },
{
"cell_type": "markdown",
- "metadata": {},
+ "metadata": {
+ "keep_output": true
+ },
"source": [
"## Latent Conditional Logit\n",
"We used a very simple MNL. Here we simulate the same MNL, by using the Conditional-Logit formulation.\\\n",
@@ -281,7 +527,7 @@
],
"metadata": {
"kernelspec": {
- "display_name": "tf_env",
+ "display_name": "choice_learn",
"language": "python",
"name": "python3"
},
@@ -295,7 +541,7 @@
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
- "version": "3.11.4"
+ "version": "3.8.18"
}
},
"nbformat": 4,
diff --git a/tests/integration_tests/models/test_latent_class.py b/tests/integration_tests/models/test_latent_class.py
index 8747ad76..1c49cf63 100644
--- a/tests/integration_tests/models/test_latent_class.py
+++ b/tests/integration_tests/models/test_latent_class.py
@@ -2,28 +2,32 @@
import tensorflow as tf
-from choice_learn.datasets import load_electricity
-from choice_learn.models.latent_class_base_model import BaseLatentClassModel
-from choice_learn.models.latent_class_mnl import LatentClassConditionalLogit, LatentClassSimpleMNL
-from choice_learn.models.simple_mnl import SimpleMNL
+tf.config.run_functions_eagerly(True)
+
+from choice_learn.datasets import load_electricity # noqa: E402
+from choice_learn.models.latent_class_base_model import BaseLatentClassModel # noqa: E402
+from choice_learn.models.latent_class_mnl import ( # noqa: E402
+ LatentClassConditionalLogit,
+ LatentClassSimpleMNL,
+)
+from choice_learn.models.simple_mnl import SimpleMNL # noqa: E402
elec_dataset = load_electricity(as_frame=False)
def test_latent_simple_mnl():
"""Test the simple latent class model fit() method."""
- tf.config.run_functions_eagerly(True)
lc_model = LatentClassSimpleMNL(
- n_latent_classes=3, fit_method="mle", optimizer="lbfgs", epochs=1000, lbfgs_tolerance=1e-8
+ n_latent_classes=2, fit_method="mle", optimizer="lbfgs", epochs=1000, lbfgs_tolerance=1e-12
)
_, _ = lc_model.fit(elec_dataset)
+ lc_model.compute_report(elec_dataset)
assert lc_model.evaluate(elec_dataset).numpy() < 1.15
def test_latent_clogit():
"""Test the conditional logit latent class model fit() method."""
- tf.config.run_functions_eagerly(True)
lc_model = LatentClassConditionalLogit(
n_latent_classes=3, fit_method="mle", optimizer="lbfgs", epochs=40, lbfgs_tolerance=1e-8
)
@@ -52,7 +56,6 @@ def test_latent_clogit():
def test_manual_lc():
"""Test manual specification of Latent Class Simple MNL model."""
- tf.config.run_functions_eagerly(True)
manual_lc = BaseLatentClassModel(
model_class=SimpleMNL,
model_parameters={"add_exit_choice": False},
@@ -70,7 +73,6 @@ def test_manual_lc():
def test_manual_lc_gd():
"""Test manual specification of Latent Class Simple MNL model with gradient descent."""
- tf.config.run_functions_eagerly(True)
manual_lc = BaseLatentClassModel(
model_class=SimpleMNL,
model_parameters={"add_exit_choice": False},