XAI MODEL DEVELOPMENT WITH PROXY APPROACH USING RIPPLE DOWN RULE METHOD

• Main Author: Mansyl, Vieri
• Format: Final Project
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/82262
• Institution: Institut Teknologi Bandung

The rapid advancement of Artificial Intelligence (AI) technology has underscored the importance of trust in AI-driven decisions. eXplainable Artificial Intelligence (XAI) systems aim to build trust by providing explanations for the decisions they make. This study aims to develop an XAI system using the Ripple Down Rules (RDR) method to construct a model capable of explaining any classification model applied to tabular data. The RDR method is implemented in the system using a proxy approach, coupled with modifications in knowledge acquisition processes involving discretization approaches. Testing of the implementation results was conducted in two phases: imitation accuracy testing to measure how accurately the RDR model mimics the original machine learning models, and explanation validity testing to assess the quality of the explanations generated by the model. The testing results demonstrate that the XAI proxy system built using the RDR method can mimic the behavior of the imitated machine learning models with an imitation accuracy rate exceeding 80% and a model explanation credibility of 89%.