基于神经网络的社保等级推荐技术研究

发布时间：2018-03-25 17:41

本文选题：社会保障　切入点：神经网络　出处：《湖南农业大学》2014年硕士论文

【摘要】：湘乡市是位于湖南省中部的一个重要工业基地和休闲旅游城市,下辖18个乡镇4个办事处,其中以普通农民占多数。在这样一个县级市中推行统筹推进“城乡社会保障体系建设”,贯彻十八大提出的“全覆盖、保基本、多层次、可持续”的12字方针,“以增强公平性、适应流动性、保证可持续性为重点”,建成覆盖城乡居民的社会保障体系是一个巨大的挑战。特别是不同的人群对社会保障体系中养老保险缴费档次存在极大的差异,如何根据个人的工作、家庭收入等特征推荐最适合的养老保险等级是一个非常值得研究的问题。针对上述问题,本文结合现代计算机技术,基于人工神经网络,利用个人信息智能推荐符合参保对象的养老保险等级,其主要工作包括： (1)基于参保人群和社保等级信息模型,从人群的地域属性、家庭属性、工作属性等多个维度出发,构建了一套社保人群多维信息体系,并对个体信息进行编码和数学建模。 (2)设计了湘乡市社保人群本体需求评价与识别特征的关系树,通过运行多维向量空间模型来表示参保对象的进行分类的特征,通过余弦定理来对特征向量的相似度进行计算,对参保对象进行研究,在保证有效分类的前提下,保证需收集的信息最少。 (3)设计了一种基于个体特征向量的神经网络分类算法,并根据现有的社保系统中的样本数据进行仿真训练和修正。实验表明,本文算法能较为准确的识别社保人群所需的养老保险费等级,相比传统的直接推荐和自主选择,有效的提高了针对性。验证了算法对参保人员分类的可行性,并研究个体信息样本对神经网络算法的影响。 (4)设计了参保人群社保等级需求分类系统框架,该框架有效地将本文提出的多维编码、参保对象需求评价树、个人信息识别特征优化、神经网络分类算法融合为一体,并分别进行了UML分析,列出了核心代码和运行界面。
[Abstract]:Xiangxiang is an important industrial base and leisure tourism city in the central part of Hunan Province, with 18 township and 4 offices under its jurisdiction. Among them, ordinary peasants account for the majority. In such a county-level city, we should carry out the "urban and rural social security system construction" as a whole and implement the 12-character policy of "full coverage, basic protection, multi-level and sustainable" proposed by the 18th CPC National Congress, "in order to enhance fairness." Adapting to mobility and ensuring sustainability is a major challenge. "building a social security system that covers urban and rural residents is a huge challenge. In particular, there are significant differences between different groups of people in terms of the level of pension insurance contributions in the social security system," he said. How to recommend the most suitable old-age insurance grade according to the characteristics of individual work and family income is a problem worth studying. In view of the above problems, this paper combines modern computer technology, based on artificial neural network, Using personal information intelligence to recommend the pension insurance grade according to the insured object, its main work includes:. 1) based on the information model of insured population and social security grade, a set of multi-dimensional information system of social security population is constructed based on the regional attribute, family attribute and job attribute of the population, and the individual information is coded and modeled by mathematics. (2) the relationship tree between the needs evaluation and identification features of social security population in Xiangxiang city is designed. The multi-dimensional vector space model is run to represent the classification features of insured objects, and the similarity of feature vectors is calculated by cosine theorem. On the premise of ensuring effective classification, the information to be collected is the least. (3) A neural network classification algorithm based on individual feature vector is designed, and the simulation training and modification are carried out according to the existing sample data of social security system. The experimental results show that, Compared with the traditional direct recommendation and independent selection, the algorithm can identify the pension insurance level required by the social security population accurately, and effectively improve the pertinence. The feasibility of the algorithm to classify the insured is verified. The influence of individual information samples on neural network algorithm is also studied. The framework of the social security classification system for insured population is designed. The framework effectively integrates the multi-dimensional coding, the object requirement evaluation tree, the feature optimization of personal information identification, and the neural network classification algorithm proposed in this paper. UML analysis is carried out, and the core code and running interface are listed.
【学位授予单位】：湖南农业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP183;TP391.3

【参考文献】