Enhancing Energy Fraud Detection with Ensemble Learning Techniques

This study aims to enhance the detection of fraudulent energy consumption patterns by integrating advanced machine learning techniques and time series analysis to improve the accuracy and robustness of classification models. Previous approaches have struggled with challenges such as missing data, imbalanced class distributions, and high-dimensional time series analysis. To address these issues, the proposed method combines several advanced techniques: Multiple Imputation by Chained Equations (MICE) for imputing missing data, AdaBoost classifiers for balancing class distributions, Piecewise Aggregate Approximation (PAA) for reducing dimensionality while preserving key patterns, and Light Gradient Boosting Machine (Light GBM) for robust classification. This integrated approach significantly improves classification performance, achieving high accuracy (94%), precision (94.5%), and recall (93.18%), and offers a scalable framework for future research and practical applications in energy management.