股权众筹项目融资成功率判别——Logistic回归与神经网络模型的比较分析

Discriminant for Financing Success Rate of Equity Crowdfunding Project:Comparative Analysis between Logistic Regression and Neural Network

基于融资项目和融资企业两个维度,构建了股权众筹项目融资成功率判别指标体系,借助探索性因子分析、非参数检验及二元Logistic回归分析对影响股权众筹项目融资的关键指标进行了筛选,着重检验了融资项目的资金注入方式、投资退出方式和融资担保条款以及融资企业的经营规模、所披露的权证资料对股权众筹融资的影响作用。在此基础上,进一步构建了神经网络模型,比较分析了回归模型与神经网络模型的判别能力。基于大伙投股权众筹平台样本数据的实证研究表明:项目维度的融资期限、视频、项目动态、项目评级和专利技术,以及企业维度的员工人数、已经营期限是判别项目融资成功率的有效指标;视频与经营性权证资料之间存在交互效应;二元Logistic回归模型的判别效果优于多层感知器模型和径向基函数模型。

In order to discriminate the financing success rate of equity crowdfunding project,this paper establishes an index system from equity crowdfunding project and enterprise.Based on the exploratory factor analysis,nonparametric test and binary logistic regression analysis,the empirical analysis identifies the key indictors which have a significant effect on project financing,especially examining the effect of the following variables,namely capital injection pattern,capital exit way,guarantee clause,enterprise scale and certificate materials.In order to guarantee the accuracy of discriminant,the paper also constructs neural network model and makes a comparative analysis between binary logistic regression model and neural network model.With the data collected from Dahuotou,the results reveal that:the key and effective indicators of project dimension are duration,videos,project updates,project ratings,patents;as well as the number of employees and business period of enterprise dimension;there also exists interaction effect between videos and operational certificate materials;binary logistic regression is superior to multilayer perceptron model and radial basis function model in discriminating the success rate of equity crowdfunding projects.