融合知识图谱与神经网络赋能数智化管理决策

Fusion of knowledge graph and neural network to empower data-intelligence for management decisions

知识图谱是感知智能通往认知智能的重要桥梁,基于知识图谱的知识表示、融合、推理将对管理决策产生深远影响.知识图谱能够有机集成符号表示和神经网络进行更有效的知识推理与决策.本文提出一套基于知识图谱的神经符号集成推理框架及三个技术方法,包括融合本体知识和生成模型的零样本决策、知识图谱嵌入表示学习增强结果异质性和可解释性、基于知识图谱的预训练模型增强面向下游任务的自动决策.该“1+3”技术体系实现了“模型驱动+知识增强”,在数智化管理决策中具有技术创新性.该体系已在实际商务管理实践中落地与应用,充分彰显场景创新性.该体系也具备通用性,可适用于多领域多情境的管理决策.

Knowledge graph(KG)is an important bridge of linking sensor intelligence to cognitive intelligence.Utilizing KG for knowledge representation,fusion and reasoning will exert great impacts on management decisions.KG helps the organic fusion of symbolic representation and neural network for more effective knowledge reasoning and decision-making.This paper develops a neural symbolic integration framework,which includes three groups of technologies and methodologies,i.e.,the integration of ontology and generative models for zero-shot decisions,the KG embedding representation learning for heterogeneous and explainable results,and the KG pretraining models for decision-making automations of down-side tasks.The proposed“1+3”technical framework embodies“model driven+knowledge enhanced”and can leverage the data-intelligence for management decisions,demonstrating the technological innovativeness.These methods and techniques have been applied in real e-commerce practices,implying the contextual innovations.The whole framework also entails the generalizability and can be applied to enhance general management decisions in various domains and contexts.