海量低被引论文的科学价值评估与预测研究

High-citation and low-citation of papers are both opposite and unified universal phenomenon in science world. Although high-citation due to its powerful positive guiding role has gotten extensive attentions from domestic and overseas scholars, we can’t neglect the negative caution role of low-citation on value and influence of scientific research. However, scholars at home and abroad give few attention on low-citation theory and its application in evaluation and forecasting of scientific value. . This project takes “massive low-cited papers” in the science world as research object and plans to importantly study the following several distinguishing work: ① We will construct a mathematical model that can define fuzzy region nearby boundary dividing high-citation and low-citation. ② We prepare to construct an optimally dynamic evolution model based on a series of fitting analyses to the time-changed curves of massive low-cited papers published by different countries and disciplines, and figure out such shape values as extremum values, slopes, smoothness and drop amplitude describing dynamic evolution curve. Then these values will be composed into structural indicator to evaluate diffusion and utilization level of literature in different countries and disciplines. ③We plan to construct a comprehensive evaluation indicator system considering a series of negative evaluation indicators and construct a new automatic evaluation model based on radar map and its area calculation theory through combining and improving traditional evaluation models. Then apply constructed indicators and model into evaluating and comparing academic quality and influence of different journals and scholars from different disciplines, and make some recommendations as to how to improve evaluation indicators and results. ④ We are ready to construct a probability forecasting model through introducing multiple condition variables based on a series of analyses and verifications on the key influencing factors leading to low-citation of papers, and then carry out a prediction analysis to future citation chance of current low-cited papers and propose some proposals improving quality and citation chance of papers.

论文高被引和低被引是科学界普遍存在且对立统一的两类现象。前者因其强大的正面导向作用受到国内外学者广泛关注，而后者对科学价值和影响力的反面警示作用不容忽视。然而国内外学者对低被引理论及其应用方面的关注相对较少，本项目以“海量低被引论文”为研究对象，拟重点完成以下几项创新性工作：① 构建一个能够界定出论文高低被引分界线附近模糊地带的数学模型。② 在拟合分析不同国家和学科海量低被引论文动态演化曲线的基础上，构建最优拟合模型测算出曲线的极值、斜率、平滑性和下降幅度等值，组成结构指标，评估不同国家和学科文献传播与利用水平。③ 引入负面评价指标，构建综合指标体系，并融合传统评价模型，构建一种新的基于雷达图及其面积求解理论的自动化评估模型，分别对不同学科期刊和学者进行评估，提出改进建议。④ 通过影响因素分析引入多条件变量，构建单篇论文价值的概率预测模型，预测当前低被引论文的未来被引机会，提出改进建议。

科技文献低被引是科学界普遍存在的现象，对科学价值和影响力的反面警示作用不容忽视。为了评估不同科研主体研发实力和影响力，本项目主要以科技文献低被引影响因素、演变规律及其评价应用为研究对象，重点完成了以下研究内容，并取得一些重要研究结果：.（1）设计和构建了三参数负指数模型，并实证分析了不同学科期刊论文零被引率随引用周期长短变化的演变模式，以及期刊论文出版后当年、三年和六年引用时间窗口零被引率随出版年不同的年度变化模式。研究发现：① 三参数负指数模型的拟合优度（R2）都达到90%以上；② 不同学科期刊论文在出版后不同长短引用时间窗口的零被引率演变曲线主要呈现出“开始下降较快，随后非常缓慢地下降，直至一个非常低且稳定的水平线”特征的凹形或偏凹形曲线。.（2）绘制出国内6个学科18本期刊文献低被引率随引用窗口变化的动态演变曲线，并测度出下降幅度、平均零被引率等指标评估不同学科期刊的传播与利用水平，期刊质量和影响力。研究发现：各学科排名靠前的6本期刊按出版后较长引用周期之后的15-19年零被引率总平均值排名，由高到低分别为航空学报、中国图书馆学报、外语教学与研究、科学通报、经济研究和分析化学。.（3）综合运用数理统计、结构方程模型和面板数据模型，全面拟合分析了科学文献低被引现象的产生原因，零被引率的各类影响因素及其之间的影响程度和路径依赖关系。研究发现：① “论文内容主题”和“期刊学术地位”是“论文零被引”的两类主要影响因素，其对“论文零被引”的影响远高于“论文个体特征”和“宣传与推荐”两类因素； ② 期刊影响因子、期刊年龄、篇均作者数、篇均页数、篇均参考文献量和期刊期数与期刊论文出版后当年和三年零被引率之间存在反向相关关系。.（4）利用社会网络分析方法识别出中国、美国和英国四个学科被引文献作者和未被引文献作者形成的科研社区差异。结果发现：美国和英国四个学科的被引作者和未被引作者形成的科研社区结构基本一致，而中国相反。