Intelligent optimization of transfer learning and knowledge tracking in enterprise strategic decision support systems

Abstract

Amidst the burgeoning advances in deep learning and artificial intelligence, strategic decision support systems for enterprises are witnessing transformative shifts. Leveraging sophisticated techniques in transfer learning and knowledge tracking algorithms, the precision and efficacy of managing enterprise risks are markedly enhanced. Consequently, this study introduces the T-DKVMN framework, an integration of Dynamic Memory Network (DKVMN) with transfer learning aimed at refining intelligent decision support systems for enterprises, particularly addressing challenges in risk identification and decision-making support. Initially, the framework harnesses pre-existing databases and labels for model pre-training, subsequently transferring the parameters from the pre-trained DKVMN model to the designated domain task via the transfer learning mechanism. This is followed by real-time updates and optimization of risk assessments using dynamic knowledge tracking. Post the pre-training and parameter transfer, the framework progresses to model training utilizing bespoke datasets and specific target labels, thus facilitating precise identification and management of enterprise risks. Experimental findings demonstrate the robust performance of the T-DKVMN framework across both public datasets and practical deployments, with risk assessment accuracy surpassing prevalent risk management methodologies. These results offer vital technical insights and a benchmark for future enhancements in intelligent enterprise decision support systems.