Original title: A PSO Based Method to Generate Actionable Counterfactuals for High Dimensional Data
Authors: Shashank Shekhar, Asif Salim, Adesh Bansode, Vivaswan Jinturkar, Anirudha Nayak
This article introduces a method using particle swarm optimization (PSO) to generate actionable counterfactual explanations (CFE) in machine learning. CFEs provide alternate predictions with minimal feature changes, helping users understand why models make certain decisions, like loan rejections. The PSO-based approach optimizes instance-centric CF generation efficiently, allowing multi-objective optimization in high dimensions and controlling attribute constraints. This method prioritizes proximity and sparsity in the generated CFEs. Evaluation against real-world datasets showcases its superior performance in action-ability metrics compared to existing methods.
Original article: https://arxiv.org/abs/2311.12825