Spatio-temporal Linear Models With Unknown Grouped Lagged Spatial And Temporal Effects And Research On The Problem Of Reward-based Crowd-funding

This paper is divided into two parts,which the first part focus on the spatial-temporal problems and the second part pay attention to the Crowd-funding problems.The two parts are independent.First Part:In this part,we propose a new model named as the spatio-temporal linear model with unknown grouped lagged spatial and temporal effects(STLM-gl),in which the spatial and temporal weight matrices are lower triangular.An algorithm STENOLS(Spatio-Temporal Elastic Net and Ordinary Least Squares)is given for estimating the model parameters.In literature,the spatial weight matrix has been given subjectively in some spatial models including spatial autoregressive models and spatial errors models,which describes the spatial structure objectively.In this part,a new method is proposed for estimating both spatial and temporal weight matrices,and can help the researchers find the hidden relationship between the observations.By using the elastic net(Enet)to perform the model selection and the ordinary least squares(OLS)to estimate the model parameters,these estimates are less biased than those obtained by only using Enet.The grouped lagged spatial and temporal structures allow us to put the same weights on some closely lagged observations in space or time,which makes sense in the realistic situations including modeling of 1980s' American housing market based on the data set containing 5543 housing price observations during 1985-1993 from Baton Rouge used in Pace et al.(2000),in which the observations are ordered chronologically.Second Part:With the rapid development of the Crowd-funding,the rapid increase in the number of projects,making investors in the project to spend a lot of time and effort.This part aims to help investors with the least time cost to select high-quality Crowd-funding projects.Under the assumption that there is a positive correlation between the quality of public projects and the completion ratio of financing(Ratio),the model in this part is modeled using the CART regression tree algorithm based on the JD Crowd-funding data.The results show that investors should focus on the Target Amount(TA),the Follower,the Progress and the Topic.The results of this part are only applicable to reward-based crowd-funding projects.For other types of Crowd-funding,the independent variables should be re-selected for model building,but the decision tree model can still be applied.