期刊h指数的集成预测研究

Integration Forecast of Journal h-index

期刊学术影响力的预测逐渐受到期刊界和学术界的广泛关注。Hirsch曾指出,相比于其他文献计量指标,h指数具有更好的预测能力,预测期刊h指数的未来发展相当于预测期刊影响力的未来演化。本文以中文社会科学引文索引为数据源库,以我国图书情报学科的13种核心期刊作为研究对象,分别建立向量自回归、向量误差修正和长短期记忆神经网络的时间序列预测模型,动态预测期刊的未来h指数。根据集成预测方法,形成上述3个模型的集成预测值,并比较各模型和方法的精度。实证结果表明,集成预测方法下的平均绝对百分比误差与均方根误差均小于3个单一的预测模型;同时,提升了预测稳定性,期刊h指数在未来呈现稳定增长趋势,图书情报领域的期刊学术影响力将保持良好的正向发展。

The prediction of academic influence of journals has gradually attracted extensive attention in journal and academic circles.Hirsch pointed out that h-index has better predictive ability compared with other bibliometric indicators.Predicting the development of journal h-index is equivalent to predicting the evolution of journal impact.On the basis of the Chinese Social Science Citation Index(CSSCI)and 13 core journals of library and information science in China,the time series prediction models of Vector Autoregression(VAR),Vector Error Correction(VEC),and Long Short-Term Memory(LSTM)are established to dynamically predict the future h-index of journals.Then,on the basis of the integrated forecast method,the integrated forecast values of the above three models are formed,and the precision of each model and method is compared.Empirical results show that the Mean Absolute Percentage Error(MAPE)and Root Mean Square Error(RMSE)of the integrated forecast method are lower than those of the three single prediction models,thereby improving prediction stability.The journal h-index shows a steady growth trend in the future,and the academic influence of the journal in the field of library and information will develop positively.