数据故事化方法:析出、重组与叙事

Data Storytelling Method:Extraction,Reorganization and Narrative

[目的/意义]数据驱动时代面临数据认知困难、解释结果晦涩难懂以及模型决策可信度不足等诸多挑战。融合可解释性结果的数据故事化方法,为应对上述挑战、增强数据利用价值提供理论支撑和解决方案。[方法/过程]梳理模型无关局部可解释性技术的解释形式、数据故事的叙事结构以及目前数据故事化研究中采用的方法,基于可解释性理论与数据故事化实现模式构建“析出—重组—叙事”的数据故事化模型,利用定义的要素元组给出数据故事映射流程,明确实现故事化模型设计的关键技术。[结果/结论]在数据故事化模型设计的理论指导下,提出面向解释结果的“扇形”故事化实现路径和融合解释结果与故事化模型要素的交互框架,并通过案例研究验证数据故事化方法在结果解释方面的实用价值。通过构建基于可解释性结果的数据故事化方法体系框架,为扩展具备数据感知与认知、可辅助智能决策功能的故事化路径提供新思路。

[Purpose/Significance]The data-driven era faces many challenges such as difficulties in data cognition,obscure interpretation results,and insufficient credibility of model decision-making.The data storytelling method that integrates interpretable results provides theoretical support and solutions to address the challenges and enhance the value of data utilization.[Method/Process]This paper summarizes the interpretation form of model-agnostic local interpretability technology,the narrative structure of data stories and the methods used in the current research on data storytelling.Based on the interpretability theory and the realization mode of data storytelling,a data storytelling model of“extraction-reorganization-narrative”is constructed,and the data story mapping process is given by using the defined element tuple.The key techniques of story model design are introduced briefly.[Result/Conclusion]Based on the theory of data storytelling model design,this paper proposes a“fan-shaped”storytelling implementation path for interpretation results and an interactive framework that integrates the elements of interpretation results and storytelling model,and reflects the practical value of data storytelling method in result interpretation through case studies.A framework of data storytelling methods based on interpretable results is constructed,which provides new ideas for expanding storytelling paths with data perception and cognition and assisting intelligent decision-making.