【作者】 张晓佳；

【导师】 吴清烈；

【作者基本信息】 东南大学 ， 管理科学与工程， 2019， 硕士

【摘要】 为了有效地提高出借人投资效率、满足出借人对个性化的投资需求,本文系统性地提出了P2P网贷平台面向出借人投资的个性化推荐系统、模型和算法,以辅助出借人进行投资决策,具体研究内容如下:本文首先通过整理历史相关文献发现:大部分学者均从理论层面对推荐系统架构、技术问题进行研究,对P2P网贷平台上出借人投资决策的具体场景则考虑较少;另外,传统个性化推荐系统、模型和算法在应用时面临多重局限性:(1)网贷产品的个性化推荐实际是投资的个性化推荐研究,与其他领域的个性化推荐区别较大。(2)出借人投资决策行为的复杂性与网贷产品属性的特殊性。然后,研究了P2P网贷平台及其个性化推荐现状,指出平台个性化功能和数据可用性与完整性较差、平台对潜在用户关注程度较低等问题;认为传统的人工推荐方法机械化、准确度不高且低效率、平台现有的服务方式无法满足海量客户对个性化的强烈需求;于是从平台的运营驱动和出借人的决策驱动两个方面确定了P2P网贷平台个性化推荐面向出借人的具体需求场景。随后,构建了面向P2P网贷出借人投资的个性化推荐系统和模型。本文以通用个性化推荐框架和模型为基础,结合出借人投资的实际场景,从三个方面进行改进:结合投资组合优化策略模块、引入动态反馈机制、增加P2P网贷平台的参与,并依据该思路对改进后的系统和模型进行详细的阐述说明。最后,对传统算法运用于P2P网贷平台的局限性进行总结,以协同过滤为基础算法,结合出借人具体的投资需求,提出了适用于P2P网贷平台出借人投资决策场景下的个性化推荐算法;并给出了一个出借人投资实例,对算法的具体步骤进行模拟并对结果进行评析,验证了该算法的有效性与可行性。更多还原

【Abstract】 In order to effectively improve the investment efficiency of lenders and meet the demand of lenders for personalized investment,this thesis systematically proposes a personalized recommendation system,model and algorithm for the lenders of P2P lending platform to assist lenders in making investment decisions,the specific research content is as follows:This thesis firstly sorts out the historical related literatures: most scholars are studying the recommendation system architecture and technical problems from the theoretical level,and consider less specific scenarios of lender investment decisions on the P2P lending platform.In addition,traditional personalized recommendations The system,model and algorithm face multiple limitations in application:(1)The personalized recommendation of online loan products is actually the personalized recommendation research of investment,which is quite different from the personalized recommendation in other fields.(2)The complexity of lender’s investment decision-making behavior and the particularity of online loan product attributes.Then,the P2P lending platform and its personalized recommendation status are studied,and the problems of platform personalization function and data availability and integrity are poor,and the platform pays less attention to potential users.The traditional manual recommendation method is considered to be mechanized and accurate.The high and low efficiency,the platform’s existing service methods can not meet the strong demand of personalized customers;so from the platform’s operation drive and the lender’s decision-driven two aspects to determine the P2P lending platform personalized recommendation for lenders Specific needs scenario.Subsequently,a personalized recommendation system and model for P2P lender investment were proposed.Based on the generalized recommendation framework and model,this paper combines the actual scenarios of lender investment and improves from three aspects: combining portfolio optimization strategy module,introducing dynamic feedback mechanism,increasing P2P lending platform participation,and based on The idea elaborates on the improved system and model.Finally,this paper summarizes the limitations of traditional algorithms applied to P2P lending platform,and uses collaborative filtering as the basis algorithm,combined with the specific investment needs of lenders,proposes personalizedization for P2P lending platform lenders’ investment decision-making scenarios.The recommendation algorithm is given.An example of lender investment is given.The specific steps of the algorithm are simulated and the results are evaluated.The validity and feasibility of the algorithm are verified.更多还原