摘要
随着科学技术的飞快发展,如何通过已有某主题词历史的研究情况预测未来该某主题词以此来判断某领域是否具有研究意义的问题已经变得越来越不可忽视.在本文中,利用主题词论文的发表数量来体现该主题词的研究热度.以往科技文献的预测模型没有统一的评判标准,导致结果的信服程度低;只是简单的考虑到某一时间范围均匀地减少,没有对具体到季节变化进行模拟,模型的准确度低;模型最优解参数难以确定,迭代周期长.本文以预测科技文献的发表数量为目的,运用时间序列预测的研究方法.通过分析前人的对于时间序列预测所题出的模型,根据科技文献的发表特点提出使用SARIMA模型进行预测.使用Python进行仿真实验,采用AIC作为评判模型的指标,采用图像初步判断模型参数范围与迭代进一步确定模型参数的办法,结果表明采用SARIMA模型得到的AIC指标为505.859相比于仅采用ARIMA模型得到AIC指标646.363表现良好.最后尝试将SARIMA模型拓展应用于其他领域,结果都表明SARIMA模型优于ARIMA模型.最后本文结论,在预测季节性数据的情况下,可以采用本文提出的实验过程,广泛应用在其他领域,表现出色同时具有实际意义.
With the rapid development of science and technology,the number of scientific and technological documents is also rapidly increasing.The question of how to predict the number of scientific and technological documents published in the field in the future based on the number of scientific and technological documents in a certain field has changed.More and more not to be ignored.In the past,the prediction models of scientific and technological literature did not have a unified evaluation standard,resulting in a low degree of conviction;it simply considered that a certain time range was uniformly reduced,and did not simulate specific seasonal changes,so the accuracy of the model was low;the model was optimal The solution parameters are difficult to determine and the iteration cycle is long.This article aims to predict the number of scientific and technological literature published,using the research method of time series forecasting.Based on the analysis of previous models for time series forecasting,according to the publication characteristics of scientific and technological literature,it is proposed to use the SARIMA model for forecasting.Use Python to perform simulation experiments,use AIC as the indicator of the evaluation model,use the image to determine the range of model parameters and iteratively to further determine the model parameters.The results show that the AIC index obtained by the SARIMA model is 505.859 compared to the AIC obtained by only the ARIMA model The indicator 646.363 performed well.Finally,I tried to extend the SARIMA model to other fields,and the results showed that the SARIMA model is better than the ARIMA model.In addition,this paper proposes the SARIMA+GARCH model,which can reduce the error caused by the SARIMA model ignoring heteroscedasticity,and further improve the accuracy of the model.Finally,the conclusion of this article is that in the case of forecasting seasonal data,the experimental process proposed in this article can be used and widely used in other fields,with outstanding p
作者
陈煜杰
王一蒙
任飞亮
CHEN Yu-jie;WANG Yi-meng;REN Fei-liang(School of Computer Science and Engineering,Northeastern University,Shengyang 110819,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2022年第10期2022-2027,共6页
Journal of Chinese Computer Systems
基金
国家级大学生创新创业训练计划项目(S202010145177)资助
中央高校基本科研业务专项项目(N182410001)资助.
