基于BP神经网络的学术论文评价模型研究

Research on Academic Paper Evaluation Model Based on BP Neural Network

[目的/意义]学术论文是学者科研水平与学术贡献的重要佐证和标志。构建科学的学术论文评价模型,对人才评价、科研经费分配、评奖评优、晋升及招聘等都具有重要指导意义。[方法/过程]文章选择Web of Science数据库中“Information Science and Library Science”学科类别下发表于2010年的论文作为研究对象。首先,基于论文多方面的关联特征构建模型特征空间;然后,采用机器学习中被广泛用于预测任务的有监督学习算法——BP神经网络训练模型,并进行十折交叉验证确保模型稳定性;最后,通过计算模型的校正决定系数(R_(adjusted)^(2))和均方根误差(RMSE),选择出最优模型。[结果/结论]本研究构建的最优BP神经网络模型的校正决定系数(R_(adjusted)^(2))达0.91,均方根误差(RMSE)约19.8,评价性能较好。

[Purpose/Significance]Academic papers serve as crucial evidence and indicators of scholars'research level and academic contributions.Constructing a scientific academic paper evaluation model holds significant guiding implications for talent assessment,research fund allocation,awards,promotions,recruitments,and more.[Methods/Process]This study selected papers published in the“Information Science and Library Science”discipline category in the Web of Science database in the year 2010 as the research subjects.Firstly,a model feature space was constructed based on various correlated characteristics of the papers.Then,a supervised learning algorithm widely used in predictive tasks in machine learning—the BP neural network—was employed to train the model.Ten-fold cross-validation was conducted to ensure model stability.Finally,by calculating the model's adjusted coefficient of determination(R_(adjusted)^(2))and root mean square error(RMSE),the optimal model was selected.[Results/Conclusion]The optimal BP neural network model constructed in this study achieved an adjusted coefficient of determination(R_(adjusted)^(2))of 091 and an RMSE of approximately 19.8,demonstrating good evaluation performance.