基于支持向量机和改进粒子群算法的科学前沿预测模型研究

Scientific Frontier Prediction Model Based on Support Vector Machine and Improved Particle Swarm Optimization

【目的/意义】识别科学研究前沿主题,预测未来发展趋势。【方法/过程】首先利用主题概率模型识别出论文、专利及基金项目多源数据中的科学前沿主题;考虑研究前沿主题演化的复杂性和非线性的特点,利用机器学习算法和支持向量机模型预测主题发展趋势,并采用改进后的粒子群算法对模型参数进行优化,以提高传统支持向量机模型在处理非线性、小样本等数据上的预测准确度。【结果/结论】实验对比发现,本方法对于研究前沿主题的预测准确度较高,准确识别出石墨烯领域未来发展趋势。

【Purpose/significance】Identifying frontier topics of scientific research and predicting future trends【Method/process】First,we use the topic probability model to identify frontier topics in the multi-source data of the paper,patent and fund projects.Considering the complexity and nonlinear characteristics of the evolution of the frontiers,we use the machine learning algorithm and the support vector machine model to predict the topic development trend,and optimize the model parameters by the improved particle swarm optimization algorithm,in order to improve the accuracy of the traditional support vector machine model on the data of the local nonlinear and small sample.【Result/conclusion】The experimental comparison shows that this method has a high accuracy in forecasting frontier topics and identifies the future development trend of graphene field accurately.