[目的/意义]新兴研究主题是科技前沿孕育与知识扩散的重要载体，其早期识别能为科技战略布局与产业决策提供支撑。文章融合多源异构数据构建早期识别与不确定性测度一体化框架，旨在提升识别准确性与稳定性，分析不确定性，为前沿监测与战略研判提供量化依据。[方法/过程]在“信息感知—认知加工—证据融合”框架下，提出一种基于神经网络与DS证据理论的多源数据融合方法。首先从多源数据中抽取主题，构建综合新兴度指标，提取特征输入BP神经网络获取后验概率，将其转化为基本概率分配，并通过DS证据理论实现多源决策级融合，从而识别新兴研究主题。其次，利用信任函数与似然函数构建信度区间，分析主题不确定性演化趋势。最后，以干细胞领域为例开展实证，并根据新兴研究主题的不确定性趋势提供发展策略。[结果/结论]相较于单一数据源，多源数据融合的方法能够有效提高新兴研究主题识别的准确性，并能够清晰刻画其早期不确定性变化趋势。
[Purpose/significance] Emerging research topics are important carriers for the incubation of scientific and technological frontiers and the diffusion of knowledge. Their early identification provides support for scientific and technological strategic planning and industrial decision-making. This paper integrates multi-source heterogeneous data to construct an integrated framework for early identification and uncertainty measurement, aiming to improve the accuracy and stability of identification, and to provide a quantitative basis for frontier monitoring and strategic assessment through uncertainty analysis. [Method/process] Within the framework of "information perception—cognitive processing—evidence fusion", this paper proposes a multi-source data fusion method based on neural networks and DempsterShafer (DS) evidence theory. First, topics are extracted from multi-source data, a comprehensive emerging-degree indicator is constructed, and features are input into a BP neural network to obtain posterior probabilities. These probabilities are then transformed into basic probability assignments, and multi-source decision-level fusion is achieved through DS evidence theory, thereby identifying emerging research topics. Second, belief intervals are constructed using belief functions and plausibility functions to analyze the evolutionary trends of topic uncertainty. Finally, an empirical study is conducted in the field of stem cells, and development strategies are proposed according to the uncertainty trends of emerging research topics. [Result/conclusion] Compared with single data sources, the multi-source data fusion method can effectively improve the accuracy of identifying emerging research topics and clearly characterize the changing trends of uncertainty in their early stages.
