赣南客家群体演化建模及成因机理研究

The carry forward of the Soviet Union is the most important factor in the revitalization of the Gannan region. The project was intended to study the basic situation of the Hakka group and its social formalization, and then to analyze its social development trajectory according to formation mechanism and evolution based on Gannan Hakka group. Through the traditional culture research, geographic information system and other technologies, we have completed the project pre-research, sorting out the Hakka migration, settlement of the source and current residence, including the name, geographical environment and other basic information, we got the following assumptions: (1) The main factor of migration is the geographical environment; (2) Hakka words is the most obvious characteristics of assimilation after settlement..In this project, firstly, we take the dynamics of group from the prior information such as topology and attribute, and construct cultural factors and propose the “scenario operator” to simulate the influence of the geographical environment. Secondly, we study the dynamics of individuals from group and propose “memory operator” to explore the function of subjectivity of individual on the formation of group, and trace the historical evolution including the spatial and temporal scales. Finally, we design the group intelligence machine to measure how the individuals move between different groups and connect with social phenomenon, and model evolution from group, realize the inverse evolution from group to topology, reveal the necessary conditions for the red cultural transmission, and provide reference for its continuation and heritage.

赣南苏区振兴若干意见中弘扬苏区精神是重中之重，本项目以赣南客家群体为研究对象，深入研究其基本状况及反映的社会形态，进而分析其形成机理与演化所反映的社会发展轨迹。通过传统文化研究、地理信息系统等技术已完成部分预研，整理收集了客家人迁移、定居的源地和现住地，包括名称、地理环境等基本资料，得出假设：(1)该群体迁移的主要因素是地理环境；(2)该群体定居后客家话是与当地文化同化中最明显的特点。.本项目首先从拓扑结构、属性等先验信息分析群体动态特性，构造文化因子，提出“情景算子”模拟地理环境等对群体演化的影响；其次从群体探究个体动态特征，提出“记忆算子”分析个体的主观性对群体形成的作用，引入时空尺度追踪历史演变；最后设计群体智慧感知器，测度个体如何在不同群体之间移动，并与社会现象关联，用群体演化建模起源，实现从群体探究拓扑等信息的反向演化，揭示苏区红色文化传播条件，为其延续和传承提供借鉴。

赣南客家群体数据作为构建异质网络的新型数据引起了广泛关注，其包含的类别信息获取代价高，且无法有效融合。这些问题导致了同类型异质网络在聚类或分类应用中性能较弱，本研究主要解决了上面2个问题，成果如下：. （1）设计了多视角图结构聚类模型。鉴于赣南客家群体数据的异质性，通过将节点属性和各类先验信息融合，设计新的边函数改进模块度分析策略，使得属性权重能自动分配，并有效结合拓扑信息，同时设计新的相似性测度方法，寻找异质网络中的有效模式。. （2）建立了基于半监督学习机制的异质网络节点分类模型，通过两层学习策略选择和找到有判别能力的属性，然后设计增强策略无缝融合各类信息，降低有效信息间的冗余度，同时增强社区结构强度，最后给出了融合的统一框架，用SVM作为基准分类器对节点分类，可以有效挖掘异质网络中的判别信息，且能作为群体感知器对动态异质网络进行演化分析。. （3）为了验证提出模型的应用前景，设计了7款软件，用于故障检测、串口分析、电路分析、间歇性故障评估和语音识别等各种场景，并获得了相应的软件著作权。同时将群体感知器用于单通道语音识别、参数决策选择和机器人路径规划等场景，申请了3项发明专利，其中2项已经授权。. （4）项目中的各类理论成果发表了期刊文章12篇，其中中科院分区Top期刊两篇，国家重要期刊1篇，满足发表高质量文章的要求。. 本项目已实现赣南客家群体演化模型，从实验上分析了其正反演化机制，在各类应用方面的成果证明了本项目具有较高的实用价值。