{"id":29132,"date":"2025-08-03T05:54:24","date_gmt":"2025-08-03T05:54:24","guid":{"rendered":"https:\/\/youzum.net\/how-to-use-the-shap-iq-package-to-uncover-and-visualize-feature-interactions-in-machine-learning-models-using-shapley-interaction-indices-sii\/"},"modified":"2025-08-03T05:54:24","modified_gmt":"2025-08-03T05:54:24","slug":"how-to-use-the-shap-iq-package-to-uncover-and-visualize-feature-interactions-in-machine-learning-models-using-shapley-interaction-indices-sii","status":"publish","type":"post","link":"https:\/\/youzum.net\/zh\/how-to-use-the-shap-iq-package-to-uncover-and-visualize-feature-interactions-in-machine-learning-models-using-shapley-interaction-indices-sii\/","title":{"rendered":"How to Use the SHAP-IQ Package to Uncover and Visualize Feature Interactions in Machine Learning Models Using Shapley Interaction Indices (SII)"},"content":{"rendered":"

In this tutorial, we explore how to use the SHAP-IQ package to uncover and visualize feature interactions in machine learning models using Shapley Interaction Indices (SII), building on the foundation of traditional Shapley values.<\/p>\n

Shapley values are great for explaining individual feature contributions in AI models but fail to capture feature interactions. Shapley interactions go a step further by separating individual effects from interactions, offering deeper insights\u2014like how longitude and latitude together influence house prices. In this tutorial, we\u2019ll get started with the shapiq package to compute and explore these Shapley interactions for any model. Check out the\u00a0Full Codes here<\/a><\/strong><\/p>\n

Installing the dependencies<\/strong><\/h3>\n
\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
!pip install shapiq overrides scikit-learn pandas numpy<\/code><\/pre>\n<\/div>\n<\/div>\n
Data Loading and Pre-processing<\/strong><\/h3>\n
In this tutorial, we\u2019ll use the Bike Sharing dataset from OpenML. After loading the data, we\u2019ll split it into training and testing sets to prepare it for model training and evaluation. Check out the\u00a0Full Codes here<\/a><\/strong><\/p>\n
\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
import shapiq\nfrom sklearn.ensemble import RandomForestRegressor\nfrom sklearn.metrics import mean_absolute_error, mean_squared_error, r2_score\nfrom sklearn.model_selection import train_test_split\nimport numpy as np\n\n# Load data\nX, y = shapiq.load_bike_sharing(to_numpy=True)\n\n# Split into training and testing\nX_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)<\/code><\/pre>\n<\/div>\n<\/div>\n
Model Training and Performance Evaluation<\/strong><\/h3>\n
\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
# Train model\nmodel = RandomForestRegressor()\nmodel.fit(X_train, y_train)\n\n# Predict\ny_pred = model.predict(X_test)\n\n# Evaluate\nmae = mean_absolute_error(y_test, y_pred)\nrmse = np.sqrt(mean_squared_error(y_test, y_pred))\nr2 = r2_score(y_test, y_pred)\n\nprint(f\"R\u00b2 Score: {r2:.4f}\")\nprint(f\"Mean Absolute Error: {mae:.4f}\")\nprint(f\"Root Mean Squared Error: {rmse:.4f}\")<\/code><\/pre>\n<\/div>\n<\/div>\n
Setting up an Explainer<\/strong><\/h3>\n
We set up a TabularExplainer using the shapiq package to compute Shapley interaction values based on the k-SII (k-order Shapley Interaction Index) method. By specifying max_order=4, we allow the explainer to consider interactions of up to 4 features simultaneously, enabling deeper insights into how groups of features collectively impact model predictions. Check out the\u00a0Full Codes here<\/a><\/strong><\/p>\n
\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
# set up an explainer with k-SII interaction values up to order 4\nexplainer = shapiq.TabularExplainer(\n    model=model,\n    data=X,\n    index=\"k-SII\",\n    max_order=4\n)<\/code><\/pre>\n<\/div>\n<\/div>\n
Explaining a Local Instance<\/strong><\/h3>\n
We select a specific test instance (index 100) to generate local explanations. The code prints the true and predicted values for this instance, followed by a breakdown of its feature values. This helps us understand the exact inputs passed to the model and sets the context for interpreting the Shapley interaction explanations that follow. Check out the\u00a0Full Codes here<\/a><\/strong><\/p>\n
\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
from tqdm.asyncio import tqdm\n# create explanations for different orders\nfeature_names = list(df[0].columns)  # get the feature names\nn_features = len(feature_names)\n\n# select a local instance to be explained\ninstance_id = 100\nx_explain = X_test[instance_id]\ny_true = y_test[instance_id]\ny_pred = model.predict(x_explain.reshape(1, -1))[0]\nprint(f\"Instance {instance_id}, True Value: {y_true}, Predicted Value: {y_pred}\")\nfor i, feature in enumerate(feature_names):\n    print(f\"{feature}: {x_explain[i]}\")<\/code><\/pre>\n<\/div>\n<\/div>\n
Analyzing Interaction Values<\/strong><\/h3>\n
We use the explainer.explain() method to compute Shapley interaction values for a specific data instance (X[100]) with a budget of 256 model evaluations. This returns an InteractionValues object, which captures how individual features and their combinations influence the model\u2019s output. The max_order=4 means we consider interactions involving up to 4 features. Check out the\u00a0Full Codes here<\/a><\/strong><\/p>\n
\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
interaction_values = explainer.explain(X[100], budget=256)\n# analyse interaction values\nprint(interaction_values)<\/code><\/pre>\n<\/div>\n<\/div>\n
First-Order Interaction Values<\/strong><\/h3>\n
To keep things simple, we compute first-order interaction values\u2014i.e., standard Shapley values that capture only individual feature contributions (no interactions).<\/p>\n
By setting max_order=1 in the TreeExplainer, we\u2019re saying:<\/strong><\/p>\n
\u201cTell me how much each feature individually contributes to the prediction, without considering any interaction effects.\u201d<\/p>\n
These values are known as standard Shapley values. For each feature, it estimates the average marginal contribution to the prediction across all possible permutations of feature inclusion. Check out the\u00a0Full Codes here<\/a><\/strong><\/p>\n
\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
feature_names = list(df[0].columns)\nexplainer = shapiq.TreeExplainer(model=model, max_order=1, index=\"SV\")\nsi_order = explainer.explain(x=x_explain)\nsi_order<\/code><\/pre>\n<\/div>\n<\/div>\n
Plotting a Waterfall chart<\/strong><\/h3>\n
A Waterfall chart visually breaks down a model\u2019s prediction into individual feature contributions. It starts from the baseline prediction and adds\/subtracts each feature\u2019s Shapley value to reach the final predicted output.<\/p>\n
In our case, we\u2019ll use the output of TreeExplainer with max_order=1 (i.e., individual contributions only) to visualize the contribution of each feature. Check out the\u00a0Full Codes here<\/a><\/strong><\/p>\n
\n
\n
\n
\n
<\/div>\n
<\/div>\n
<\/div>\n<\/div>\n
Copy Code<\/span>Copied<\/span>Use a different Browser<\/span><\/a><\/div>\n<\/div>\n
si_order.plot_waterfall(feature_names=feature_names, show=True)<\/code><\/pre>\n<\/div>\n<\/div>\n
\n
\"\"<\/figure>\n<\/div>\n
In our case, the baseline value (i.e., the model\u2019s expected output without any feature information) is 190.717.<\/p>\n
As we add the contributions from individual features (order-1 Shapley values), we can observe how each one pushes the prediction up or pulls it down:<\/strong><\/p>\n