基于进化计算的新闻传播应对策略优化研究

With the development of Internet, massive news is disseminated in the internet with considerably fast speed simultaneously in multiple directions. As a consequence, how to model the characteristics of news dissemination and how to optimize news dissemination to improve the effect of good news has become key and urgent issues needed to be devoted great attention. However, due to the great complexity and dynamics of news dissemination, it is difficult or even impossible to settle the issues related to the management of news dissemination by means of traditional mathematic models and optimization methods. To resolve this predicament, based on the news data, this project aims at devising novel optimization approaches to the management of news dissemination based on evolutionary algorithms which do not rely on mathematic models to solve a practical problem. Specifically, combining the classical news dissemination models with news data, we will firstly design data-driven evaluation mechanisms to assess the effect of different management strategies of news dissemination. Secondly, based on the designed evaluation mechanisms, we will devise multiple data-driven optimization objectives for the optimization of the management of news dissemination. In this way, we can transform the management optimization into a multi-objective combinational optimization problem. Thirdly, to resolve the devised multi-objective optimization problem, we will design novel large-scale multi-objective evolutionary algorithms to obtain the optimal solutions by taking account of the uncertainty of news dissemination. Further, to improve the execution efficiency of the proposed algorithms, we will further develop distributed multi-objective evolutionary algorithms based on probability distribution models, so that the proposed evolutionary algorithms could afford high-quality solutions in acceptable time. At last, we will design a prototype system via integrating the above developed data-driven assessment approaches of different management strategies and the devised optimization algorithms to realize the intelligent optimization of the management of news dissemination. By means of this project, we expect to construct quantitative analysis mechanisms for different management strategies of news dissemination, which, as far as we know, is the first try along this direction. At the same time, it is also the first time to apply evolutionary algorithms to resolve the optimization of news dissemination management. With the outcome of this project, we expect to improve the effectiveness and efficiency of the management of news dissemination and remedy the defect that traditional management of news dissemination seriously relies on the prior knowledge and human experience.

随着互联网的发展，大量新闻在网络中快速多向传播，如何刻画新闻传播特征，优化传播效果，是新媒体时代亟需解决的重要课题。由于新闻传播的复杂性，难以采用传统的建模与优化技术对新闻传播的管理与应对策略进行评价和优化。针对此，本项目基于新闻数据，结合进化计算不依赖于问题模型的优势，提出基于进化算法的新闻传播应对优化方法。首先，将网络传播模型和新闻数据相结合，提出数据驱动的传播应对策略效用评价机制；其次，基于评价机制，将新闻应对策略组合及其作用域的选择问题转化为多目标优化问题；第三，提出基于进化计算的新闻传播应对优化方法，重点针对传播过程的不确定性和时效性问题，提出基于概率分布的分布式进化算法；最终，实现原型系统验证方法的有效性。本项目是对新闻传播应对策略定量评价的新探索，也是进化计算在新闻传播领域应用的新尝试，有助于改变传统上基于新闻从业者先验知识选择传播应对手段的不足，提升传播效用。

随着互联网的发展，大量新闻在网络中快速多向传播，如何刻画新闻传播特征，优化传播效果，是新媒体时代亟需解决的重要课题。由于新闻传播的复杂性，难以采用传统的建模与优化技术对新闻传播的管理与应对策略进行评价和优化。针对此，本项目基于新闻数据，结合进化计算不依赖于问题模型的优势，提出基于进化算法的新闻传播应对优化方法。研究内容包括：1）将网络传播模型和新闻数据相结合，提出数据驱动的传播应对策略效用评价机制；2）基于评价机制，将新闻应对策略组合及其作用域的选择问题转化为多目标优化问题；3）提出基于进化计算的新闻传播应对优化方法，重点针对传播过程的不确定性和时效性问题，提出基于概率分布的分布式进化算法；4）验证方法的有效性。. 项目取得的主要研究进展包括：1）针对进化计算在大规模网络优化中的效率问题，原创性地提出了矩阵基进化计算方法，提出了大规模高维搜索空间中高可并行进化计算方法，有效地提高了进化计算在大规模网络优化中的效率和可扩展性；2）针对传播策略优化中需要兼顾公平性和满意度的多目标优化问题，提出了考虑公平性和满意度的多目标进化算法，提出了多峰值多模态场景的自适应进化计算方法，有效地应维持算法的搜索多样性并应对多目标优化的挑战；3）针对传播策略优化中目标函数评估昂贵、需要依赖数据驱动的挑战，提出了基于模型集成与数据生成的高效数据驱动优化方法，提出带隐私保护的边云协同数据驱动优化方法，实现了分布式数据驱动的高效进化优化；4）在算法的应用验证方面，将上述方法集成应用于新闻传播应对策略优化等问题，既验证了所提出的算法的有效性，又为求解相关问题提供了新的有效途径。依托项目共发表了65篇已标准基金号的学术论文，其中IEEE Trans.系列期刊论文44篇。此外，项目组还申请发明专利3项，其中已获得授权1项；依托本项目培养博士研究生11人，其中已获学位8人。. 本项目是对新闻传播应对策略定量评价的新探索，也是进化计算在新闻传播领域应用的新尝试，有助于改变传统上基于新闻从业者先验知识选择传播应对手段的不足，提升传播效用。