<![CDATA[Traditional reactive logistics auditing often fails to detect hidden operational inefficiencies, particularly in transaction data with imbalanced class distributions. This study aims to analyze the effect of Genetic Algorithm (GA)-based hyperparameter optimization on the performance of deep learning models for logistics expenditure efficiency classification and to compare its performance with baseline Artificial Neural Network (ANN) and Convolutional Neural Network (CNN) models. This study used a real-world logistics transaction dataset, with the Cost per Kilogram metric employed as the basis for classification labeling. The research stages included data collection, preprocessing and feature engineering, data balancing using the Synthetic Minority Over-sampling Technique (SMOTE), development of ANN and CNN models, ANN hyperparameter optimization using GA, and comparative evaluation based on accuracy, precision, and recall. The results showed that the baseline ANN and CNN models obtained a recall of 0.00 in detecting inefficient transactions, whereas the GA-optimized ANN achieved an accuracy of 91% and a perfect recall of 1.00. These findings indicate that GA-based hyperparameter optimization improves the model's ability to detect inefficient transactions and reduces diagnostic blind spots in imbalanced logistics financial data. This study contributes a high-accuracy classification approach that can support proactive financial monitoring and automated auditing in the logistics sector.]]>
Copyrights © 2026
