SHAP-SHAPley-Model-code

This is for educational purposes. This data is synthesized. Do not cite data for publications. Please cite author for code.

import numpy as np import pandas as pd from sklearn.model_selection import train_test_split from sklearn.ensemble import RandomForestRegressor import shap import matplotlib.pyplot as plt

1. Generate some synthetic data for demonstration

np.random.seed(86) X = pd.DataFrame({ 'Diabetes': np.random.rand(100) * 10, 'Cancer': np.random.rand(100) * 5, 'COPD': np.random.randint(0, 2, 100) }) y = 2 * X['Diabetes'] + 3 * X['Cancer'] - 5 * X['COPD'] + np.random.randn(100) * 2

2. Split data into training and testing sets

X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=86)

3. Train a machine learning model (e.g., RandomForestRegressor)

model = RandomForestRegressor(n_estimators=100, random_state=86) model.fit(X_train, y_train)

print("Model training complete. R-squared on test set: {:.2f}".format(model.score(X_test, y_test)))

4. Initialize a SHAP explainer

For tree-based models, shap.TreeExplainer is efficient

explainer = shap.TreeExplainer(model)

5. Calculate SHAP values for the test set

It's often good practice to explain predictions on a subset of the data,

especially for larger datasets, or specifically the test set.

shap_values = explainer.shap_values(X_test)

6. Visualize the SHAP scores

print("\nGenerating SHAP summary plot...")

Summary plot: illustrates the impact of each feature on the model's output

Red indicates higher feature value, blue indicates lower feature value

shap.summary_plot(shap_values, X_test, plot_type="bar", show=False) plt.title("SHAP Feature Importance (Mean Absolute SHAP Value)") plt.show()

shap.summary_plot(shap_values, X_test, show=False) plt.title("SHAP Summary Plot") plt.show()

print("\nSHAP values calculated and visualized. The plots show feature importance and individual contributions to predictions.") print("You can also inspect individual predictions, for example, the first test instance:") shap.initjs() shap.force_plot(explainer.expected_value, shap_values[0,:], X_test.iloc[0,:])