Home▸Machine Learning & Neural Networks▸Explaining a Loan Model: Exact Shapley Values Compared with Occlusion, Gradient and LIME Attributions

Explaining a Loan Model: Exact Shapley Values Compared with Occlusion, Gradient and LIME Attributions

Open the black box of a small loan-approval model. Compute exact interventional Shapley values by enumerating all 32 feature coalitions, then compare them with occlusion, gradient times input and a LIME-style local surrogate on the same applicant. See why the methods agree on a linear model and disagree when features interact. Waterfall modelled in Blender, with six numerical checks built in.

Machine Learning & Neural Networks3DModerate60 FPS📱 Mobile-adapted🟠 Blender⇄ 2D version
Explaining a Loan Model: Exact Shapley Values Compared with Occlusion, Gradient and LIME Attributions ↗ Open standalone

Open the black box of a small loan-approval model. Compute exact interventional Shapley values by enumerating all 32 feature coalitions, then compare them with occlusion, gradient times input and a LIME-style local surrogate on the same applicant. See why the methods agree on a linear model and disagree when features interact. Waterfall modelled in Blender, with six numerical checks built in.

⚙ Under the hood

Open the black box of a small loan-approval model. Compute exact interventional Shapley values by enumerating all 32 feature coalitions, then compare them with occlusion, gradient times input and a LIME-style local surrogate on the same applicant. See why the methods agree on a linear model and disagree when features interact. Waterfall modelled in Blender, with six numerical checks built in.

Three.jsAI explainabilityXAISHAPLIMEneural networkblack boxmachine learningExplainable AIShapley ValuesFeature Attribution

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install

What did you find?

Add reproduction steps (optional)