RBSTIP: A Rule-Based Framework for Balancing Accuracy and Interpretability in Predictive Modeling

Achieving a balance between predictive accuracy and interpretability remains a key challenge in the development of transparent machine learning systems. This research introduces a novel Rule-Based System for Transparent Prediction (RBSTIP), designed to enhance both the clarity and performance of predictive models. By leveraging rule-based algorithms, RBSTIP aims to provide interpretable insights into complex decision processes, enabling stakeholders to better understand the reasoning behind predictions. The proposed approach addresses the often conflicting goals of accuracy and transparency by optimizing the PerformanceInterpretability (PI) value, maintaining it at or below 0.05 to ensure practical feasibility. For experimental evaluation, we employed the widely used Income dataset, selected due to its structured, real-world characteristics and its prevalence in benchmarking predictive modeling techniques. This dataset provided a relevant and robust basis for assessing the effectiveness of RBSTIP compared to traditional models such as decision trees, linear models, and neural networks.