Explainable AI for Tree-Based Models: From SHAP to PDP and Feature Interactions

In this AI Tech Experts Webinar, Jakub Cieślak, Senior Data Scientist, shows how to interpret tree-based models using practical XAI techniques and how to avoid the most common traps. The talk focuses on tabular ML problems and walks step by step through: 👉 local vs global explanations and when to use each, 👉 SHAP values and breakdown plots for individual predictions, 👉 partial dependence plots (PDP) and ICE for global feature effects, 👉 feature importance vs causality and why this distinction matters, 👉 detecting data leakage, hidden bias and misleading patterns, 👉 common XAI mistakes and concrete recommendations. Examples are illustrated on a classic Titanic dataset, but the lessons apply directly to real-world decision systems used by product teams, stakeholders and regulated industries. 00:00 Intro & Agenda 00:30 Introduction to explainable AI 03:33 Explaining predictions with SHAP 06:27 Why explain predictions: quality and trust 09:50 Legal and ethical requirements for XAI 11:31 Common mistakes and best practices Check our website: https://deepsense.ai/ Linkedin:   / applied-ai-insider   #ExplainableAI #XAI #TreeBasedModels #ModelInterpretability #MachineLearning #DataScience #FeatureImportance #SHAP #PartialDependence #MLTransparency