论文部分内容阅读
【目的/意义】探讨学术期刊影响力指数与影响因子等传统期刊计量指标的相关性并构建该指标值的预测模型。【方法/过程】首先以图情领域19种核心期刊为研究对象,以SPSS16.0为分析工具对影响力指数与22种传统期刊计量指标的相关性进行分析,得到与之显著相关的15种传统计量指标。经主成分分析消除这15个指标间的相关性后,将其用作BP神经网络CI值预测模型的输入向量,同时采用“综合性人文、社会科学”类的632个期刊的数据作为训练样本对网络进行训练。【结果/结论】使用训练好的BP神经网络对19种图情领域核心期刊的CI值进行预测,结果显示了较高的预测精度。该模型可用于影响力指数值的预测及期刊学科内排名的预估。
[Purpose / Significance] To explore the correlation between academic journals ’influence index and influencing factors and other traditional journals’ metrological indicators and to construct a predictive model of the index value. 【Method / Process】 Firstly, 19 core journals in the field of plot and plot were selected as the research object, SPSS16.0 was used as the analysis tool to analyze the correlation between the impact index and the measurement index of 22 traditional journals, and 15 Traditional measurement index. After the principal components analysis eliminated the correlation between the 15 indicators, it was used as the input vector of the CI value prediction model of BP neural network. At the same time, the data of 632 journals of “Comprehensive Humanities and Social Sciences” Training samples to train the network. [Results / Conclusion] The CI values of 19 core journals in the field of pictorial information were predicted using the trained BP neural network, and the results showed high prediction accuracy. The model can be used for the prediction of the influence index value and the forecast of ranking within the periodicals.