An End-to-End Explainable AI Framework with Automated LLM-Based Natural Language Explanation Generation for Energy Systems

Jul 16, 2026

Upload by Ivy

Explainable AI (XAI) enables machine learning deployed in high-stakes domains by improving transparency, trust, and accountability where black-box models remain difficult to interpret. The framework combines prediction, explanation generation, evaluation, and automatic conversion of XAI outputs into natural language for non-technical stakeholders. Tabular preprocessing and deep neural networks support both classification and regression, using LIME and SHAP for local and global explanations, evaluated via fidelity and stability, then reformulated as structured inputs to an LLM. Experiments on power fault detection and building energy labeling show strong accuracy and near-perfect explanation quality.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages25

Reading Time1 h 3 min

1. Introduction

1.1 Background: XAI and Black-Box Challenges

1.2 Post-hoc Explanation Methods (LIME and SHAP)

1.3 Alternative Explainability Techniques

An End-to-End Explainable AI Framework with Automated LLM-Based Natural Language Explanation Generation for Energy Systems Venkata Sesha Sai Raj Nanduri1, Akthar Hussain2 and Van-Hai Bui 1,* 1 Department of Computer and Information Science, University of Michigan-Dearborn, 4901 Evergreen Rd, Dearborn, MI 48128, USA 2 Department of Electrical and Computer Engineering, Laval University, 2325 Rue de l’Université, Québec, QC G1V 0A6, Canada Abstract Explainable AI (XAI) is important for deploying machine learning systems in domains where stakes are very high and where transparency, trust and accountability are critical. Although black box models like deep neural networks often perform with high efficiency, interpreting their decisions remains a difficult task. This paper proposes a reusable end-to-end XAI framework that is the combination of prediction, explanation generation, evaluation and converting these explanations into natural language text of explanation which can be easily understood by the non-technical stakeholders as well. This framework initially preprocesses tabular datasets using standardized feature transformation techniques and trains deep neural network for both classification and regression tasks. Local and global explanations are generated using XAI algorithms like Local Interpretable Model-agnostic Explanation (LIME) and SHapley Additive exPlanations (SHAP) respectively. To evaluate these explanations, we use metrics like fidelity and stability to know how accurately and consistently explanations reflect the model behavior. The generated explanation includes feature importance scores, prediction specific attributes. These are then transformed into a structured input to the Large Language Model (LLM), which generates a natural language explanation through which everyone can understand the explanations generated by XAI algorithms. This framework is tested on power system fault dataset detection dataset and building energy labels dataset. The experimented results demonstrate the effectiveness of the proposed approach across both classification and regression tasks. For fault detection, the neural network model achieved 99% accuracy with ROC-AUC score of 1.00. For building energy prediction, model achieves R2 score of 0.67. Furthermore, fidelity and stability also stays at the top with a nearly perfect score which makes the model very reliable. These findings says that the proposed approach produces a stable and faithful explanations while improving the interpretability of black box model to everyone with the help of LLMs. This framework provides a model agnostic solution for enhancing transparency in both classification and regression tasks. Keywords: Explainable AI; Large language model; LIME; Neural networks; SHAP 1. Introduction Artificial Intelligence (AI) and Machine Learning (ML) have become the integral part across numerous application domains, including healthcare, finance, transportation and energy systems etc. Their ability to automatically learn the complex patterns from huge datasets has enabled significant improvements in performance and decision-making. However, as AI systems are most frequently used in the high-stakes domains where decisions directly affect the lives of people, the ability to understand and justify their predictions has become equally important. This growing demand has led to the rapid development of Explainable Artificial Intelligence (XAI), which aims to make machine learning models more transparent, interpretable and trustworthy while preserving the performance [1 – 3] . Among the ML techniques, deep neural networks are good at solving difficult problems because of their ability to learn highly nonlinear relationships from huge amount of data. Consequently, they have been widely adopted for applications such as power system fault diagnosis, computer vision, natural language processing and energy prediction where accurate and interpretable predictions are esse

FAQs

What is the main goal of the proposed XAI framework?

The framework aims to generate explanation text in natural language by combining end-to-end prediction, XAI explanation generation, and evaluation, then converting explanations for non-technical stakeholders using an LLM.

How are local and global explanations produced in the framework?

Local explanations are generated using LIME, while global explanations are generated using SHAP to describe overall model behavior.

What metrics are used to evaluate whether explanations are faithful and consistent?

Explanations are evaluated using fidelity and stability metrics to assess how accurately and consistently they reflect model behavior.

Jul 16, 2026

Upload by Ivy