【作者】 李敏；

【导师】 朱容波；

【作者基本信息】 中南民族大学 ， 计算机技术（专业学位）， 2019， 硕士

【摘要】 精准扶贫、信息化扶贫是当前我国扶贫工作的重要指导思想。互联网时代和大数据时代的到来为各行业的发展提供了机遇,同时也为扶贫工作的信息化提供了契机。当前,计算机技术应用于贫困户扶贫,主要体现在三个方面:第一,应用计算机系统有效管理扶贫信息,提高办事效率,加速扶贫进程;第二,利用大数据处理技术高效统计分析数据变化、数据指标;第三,利用机器学习技术,对贫困户数据做分类识别或预测,协助帮扶人员决策。虽然计算机技术应用于扶贫工作取得了较大进展,但仍然存在一些缺陷,导致扶贫效果迟缓。这主要体现在以下两个方面:第一,缺少从多方面、多维度精细刻画贫困户特征的方法,缺少可解释性,只回答是什么,难以回答为什么;第二,对构建人机互动反馈闭环的研究较少,只研究或实现决策至结果单一方向的方法或系统。针对上述问题,本文将用户画像技术运用到精准扶贫领域,结合精准扶贫实际业务需求和扶贫数据实际情况,构建出一套贫困户画像标签体系和贫困户画像模型,并实现了基于用户画像的精准扶贫可视化系统中贫困户画像子系统部分。本文主要工作如下:(1)在特征工程的基础上,从多角度、多维度提取贫困户本体特征,构建贫困户标签体系,从而构建多标签、多模型融合的贫困户画像模型,使得精准扶贫系统具备精细刻画贫困户的能力,同时具有可解释性。在贫困户标签体系的基础上,借助大数据处理平台,使用聚类和降维技术,研究用多模型分层融合方法将贫困户分级,提出一种提取贫困群体共同特征的方法。系统实验表明,各贫困群体之间刻画出的特征具有明显区分性和可解释性。(2)提出一种分析群体致贫、脱贫原因的方法,具有较好的可解释性、可追踪性,帮助帮扶人员认识贫困群体特征,把握扶贫效果,提供制定政策参考建议,从而构建扶贫反馈闭环。实现了贫困户脱贫轨迹的刻画,使得本系统具备追踪每个阶段扶贫成绩以及观察贫困户多方面状态变化的能力。(3)基于上述方法和技术,完成基于用户画像的精准扶贫可视化系统的整体架构设计。在此架构基础上,实现贫困户画像子系统的三大功能模块:贫困户画像模块、脱贫群体原因分析模块、脱贫轨迹展示模块。为了验证系统是否满足预期需求,应用工程测试方法对系统做了全面测试,并且提供了相关应用案例展示。测试结果表明,系统符合预期。更多还原

【Abstract】 Targeted poverty alleviation and information poverty alleviation are important guiding ideologies for poverty alleviation work in China.The advent of the Internet Era and the Era of Big data has provided opportunities for the development of various industries,and also provided an opportunity for the informationization of poverty alleviation work.After decades of development,poverty alleviation has accumulated a large amount of potentially valuable data,using big data technology to build a targeted poverty alleviation visualization system based on user profile,and tapping potential poverty alleviation information models is of great significance for achieving personalized poverty alleviation and informationization poverty alleviation.At present,the application of computer technology to poverty alleviation is mainly reflected in three aspects.First,using computer systems to effectively manage poverty alleviation information,improve work efficiency,and accelerate poverty alleviation.Second,the use of Big data processing technology to efficiently analyze statistical data changes and data indicators.Third,using machine learning technology to predict or classify poor household data to help decision-making.Although computer technology has been applied to poverty alleviation and has made great progress,there are still some shortcomings,and the poverty alleviation effect is slow.First,Lack of characterizing poor households from multiple aspects and multiple dimensions,lack of interpretability.Only answer what is,difficult to answer why.Second,there is less closed-loop research on constructing human-computer interaction feedback,only research or implementation of decision-to-result single direction methods or systems.In response to the above problems,this paper applies user profile technology to the field of targeted poverty alleviation,combined with the actual business needs of poverty alleviation,the actual situation of poverty alleviation data and user profile technology,and constructs a set of poor household profile labeling system and poor household profile model based on poverty alleviation data.In addition,achieved a targeted poverty alleviation visualization system based on user profile.The main work of this paper is as follows:(1)This paper extracts the characteristics of poor households from various aspects,builds a multi-label,multi-model fusion method of poor households on the basis of characteristic engineering,thus constructing a label system for poor households,so that the targeted poverty alleviation system has a fine depiction of poor households and is interpretable.Using clustering and dimension reduction techniques,and using the big data processing platform,based on the label system of poor households,the paper studies the hierarchical integration of multiple models to classify poor households and proposes how to extract the common characteristics of poor groups.Experiments show that the features depicted between groups are clearly distinguishable.(2)Putting forward the analysis of the causes of poverty alleviation and poverty alleviation in poverty alleviation groups,not only effectively solving the problem of “who is a poor resident”,but also effectively solving the problem of “what is the cause of poverty” and “how to help”,further analyzing poverty alleviation to build a closed loop of policy poverty alleviation feedback.Depicting the poverty alleviation track of poor households,the system has the ability to track poverty alleviation at each stage and the multi-faceted state changes of poor households.(3)Based on the above method and technology,the targeted poverty alleviation visualization system based on user profile is realized,which makes the system have the ability to describe the characteristics of poor households.It has interpretability and traceability,and there is feedback loop,which can effectively help decision makers to achieve targeted poverty alleviation.Based on engineering tests,the system was fully tested and an application case was presented.The test results show that the system is in line with expectations.更多还原