Evaluation of the Digital Transformation Effects in Manufacturing Using the DEA-BP Model and the Internet of Things

Abstract

This work aims to comprehensively evaluate the effects of digital transformation in the manufacturing industry by employing a combined approach of data envelopment analysis (DEA) and Back Propagation (BP) neural network to construct the DEA-BP model. Firstly, the digital transformation effects are more comprehensively revealed by constructing the DEA-BP model, leveraging the efficiency evaluation of DEA and the nonlinear learning capabilities of BP neural networks. Secondly, critical input factors are selected. This work considers the manufacturing environment driven by the Internet of Things (IoT) to assess the core influencing factors of digital transformation more practically and operationally. Finally, through experiments utilizing simulated manufacturing process data, the performance of various models is compared in terms of overall efficiency, prediction performance, and classification performance. The research results indicate that the DEA-BP model significantly outperforms other models in overall efficiency evaluation, reaching a maximum efficiency of 93%, fully capitalizing on the flexibility of DEA and the nonlinear learning capabilities of the BP model. Regarding prediction performance for digital transformation, the DEA-BP model exhibits higher accuracy. In classification performance, the DEA-BP model remarkably improves accuracy, precision, and recall, demonstrating higher stability than other models. This work provides a new approach to evaluating the effects of digital transformation in the manufacturing industry, offering feasibility and guidance for practical applications, and it possesses high research and application value. Future research could further optimize model interpretability and computational efficiency, explore additional evaluation indicators, and enhance comprehensiveness and applicability.