基于机器学习算法的研究热点趋势预测模型对比与分析

摘要：[目的/意義]细粒度分析学科领域热点主题发展脉络并对利用机器学习算法对未来发展趋势进行准确预测研究。[方法/过程]提出一种基于机器学习算法的研究热点趋势预测方法与分析框架，以基因工程领域为例利用主题概率模型识别WOS核心集中论文摘要数据研究热点主题并进行主题演化关联构建，然后选取BP神经网络、支持向量机及LSTM模型等3种典型机器学习算法进行预测分析，最后利用RE指标和精准度指标评价机器学习算法预测效果并对基因工程领域在医药卫生、农业食品等方面研究趋势进行分析。[结果/结论]实验表明基于LSTM模型对热点主题未来发展趋势预测准确度最高，支持向量机预测效果次之，BP神经网络预测效果较差且预测稳定性不足，同时结合专家咨询和文献调研表明本文方法可快速识别基因领域研究主题及发展趋势，可为我国学科领域大势研判和架构调整提供决策支持和参考。

关键词：热点主题;发展趋势;机器学习;LSTM模型;支持向量机模型

DOI：10.3969/j.issn.1008-0821.2019.04.003

〔中图分类号〕G203〔文献标识码〕A〔文章编号〕1008-0821（2019）04-0023-11

Comparison and Analysis of Research Trend Prediction

Models Based on Machine Learning Algorithm

——BP Neural Network，Support Vector Machine and LSTM Model

Li Jing1Xu Lulu2*

（1.School of Economics and Management，Tongji University，Shanghai 200092，China;

2.Department of Information Resources Management，Business School，Nankai University，

Tianjin 300071，China）

Abstract：[Purpose/Signficance]Fine-grained analysis of the development context of hot topics in the subject field and accurate prediction of future development trends using machine learning algorithms.[Method/Process]This paper proposed a research hotspot prediction method and analysis framework based on machine learning algorithm.Taking the field of genetic engineering as an example，it used the topic probability model to identify the hot topics of the WOS core summary data and constructed the theme evolution association.Then selected three typical machine learning algorithms，such as BP neural network，support vector machine and LSTM model to predict and analyze.Finally，the prediction results of the machine learning algorithm were evaluated by using RE index and precision index，and analysed the research trend in the fields of medicine and health and agricultural food in the field of genetic engineering.[Result/Conclusion]The experiment showed that the LSTM model had the highest prediction accuracy for the future development trend of hot topics，the prediction effect of support vector machine was the second，the prediction effect of BP neural network was poor and the prediction stability was insufficient.At the same time，combining expert consultation and literature research，it showed that this method could quickly identify the topic and development trend of gene field.It could provide decision support and reference for the judgement and adjustment of the discipline in China.

Key words：hot topics;development trend;machine learning;LSTM model;support vector