具有分组效应的时空滞后模型与奖励型众筹问题研究

发布时间：2018-05-26 02:43

本文选题：弹性网络 + 分组滞后空间结构　；参考：《中国科学技术大学》2017年硕士论文

【摘要】：本篇论文由两部分构成。第一部分研究的是时空滞后模型的相关问题,第二部分研究的是利用决策树解决众筹领域的相关问题。两部分独立进行研究,相互之间并无关联。第一部分:本部分提出了时空模型STLM-gl,这个模型的时空权重矩阵为下三角阵且具有分组效应。算法 STENOLS(Spatio-Temporal Elastic Net and Ordinary Least Squares)可以用来估计上述模型的参数。空间模型的(空间滞后模型与空间误差模型)空间权重矩阵在以往的研究中通常是主观给定,这往往与真实的空间结构有所出入。在该部分中,算法STENOLS就是用来估计模型STLM-gl中未知的时空权重矩阵,其可帮助研究人员更好的找到隐藏的时空结构。通过利用弹性网络估计时空权重矩阵中不为0的元素,再利用最小二乘法估计具体参数值,其产生的偏差比仅用弹性网络要小的多。Pace于2000年发表的论文已经表明了分组效应在实际应用场景中是具有实际意义的。第二部分:随着众筹行业的迅猛发展,众筹项目数量迅速增长,使得投资者在项目选择上花费了大量的时间精力。本部分旨在帮助投资者以最少时间成本选择优质的众筹项目。在假设众筹项目优质程度与融资完成比有正相关关系的前提下,本部分基于京东众筹数据,利用CART回归树算法进行决策树建模。研究结果表明,投资者应重点关注目标金额,关注人数,项目进展和话题这四个指标。本部分研究结果仅适用于奖励型众筹,对于其他类型众筹应当重新选择自变量进行模型建立,但决策树模型仍然可以适用。
[Abstract]:This thesis consists of two parts. In the first part, the related problems of time-space lag model are studied. In the second part, the decision tree is used to solve the related problems in the field of crowdfunding. The two parts are independent of each other and have no relationship with each other. In the first part, a spatio-temporal model STLM-glis is proposed. The space-time weight matrix of this model is a lower triangular matrix and has grouping effect. The algorithm STENOLS(Spatio-Temporal Elastic Net and Ordinary Least Squares) can be used to estimate the parameters of the above model. The spatial weight matrix of spatial model (spatial lag model and spatial error model) is usually given subjectively in previous studies, which is often different from the real spatial structure. In this part, the algorithm STENOLS is used to estimate the unknown space-time weight matrix in the model STLM-gl, which can help researchers find the hidden space-time structure better. By using the elastic network to estimate the non-zero elements in the space-time weight matrix, and then using the least square method to estimate the specific parameter values, The paper published in 2000 by .Pace, which produces a much smaller deviation than the elastic network alone, has shown that the grouping effect is of practical significance in practical application scenarios. The second part: with the rapid development of crowdfunding industry, the number of crowdfunding projects increases rapidly, which makes investors spend a lot of time and energy on project selection. This section aims to help investors select high-quality crowdfunding projects at minimum time cost. Under the assumption that there is a positive correlation between the quality of crowdfunding project and the ratio of financing completion, this part uses CART regression tree algorithm to model decision tree based on JingDong crowdfunding data. The results show that investors should focus on the target amount, number of people, project progress and topic. The results of this study are only applicable to reward crowdfunding. For other types of crowdfunding, independent variables should be re-selected for modeling, but the decision tree model can still be applied.
【学位授予单位】：中国科学技术大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F724.6;F832.4

【参考文献】