Research Of Reward Crowdfunding Based On Machine Learning

Crowdfunding,as an emerging Internet financial model,breaks the barriers of traditional financial models and provides new financing channels and low financing threshold for small,medium and micro enterprises.However,not all crowdfunding projects can be successfully funded,and the success rate of crowdfunding projects on various crowdfunding platforms is only between 40% and 80%.Therefore,how to improve the success rate of crowdfunding projects and how to apply limited resources and energy in key areas are the issues that need to be concerned and urgently considered by fund-raisers.The purpose of this study is to explore the important factors affecting the amount of crowdfunding,so as to provide theoretical support for the decision of fund-raisers.This paper analyzes and studies the factors influencing the final financing amount of crowdfunding,and constructs the financial influencing factors index system of crowdfunding financing from four aspects,the basic information dimension of the project,the behavioral information dimension of the investors,the information dimension of the fund-raisers,and the innovative addition of the publicity and interactive information dimension of the project.Using the crawler technology to get the raise relevant data and the related data of the publicity form the original data set,and the original image data in the data are analyzed in character recognition,text and topic extraction steps such as summary formed the raise project text data related indicators,and through missing value processing,unbalanced data sets processing process so as to obtain the training data set,using machine learning model and deep learning model for training data set,and use the training after the model to predict the raise project financing amount,in addition,according to the result of tree model calculating weight weighted by various factors affecting the weight,Combined with the correlation analysis results of various influencing factors,the factors affecting the final funding amount of the crowdfunding project and the weight of each factor are screened.Through this paper,the study found that project basic information,investor behavior,information,financing and project publicity and interaction information four aspects are the result of the project financing has a certain degree of influence,among them,the number of supporters,pay attention to the number of projects,and project sites like number,financing support number and propaganda weibo several factors,such as raising money for the most significant influence the project raise a number of factors,and the raise project description language style,is there a product introduction video to all the factors,such as raising money to raise the influence degree of the weak.Therefore,fundraisers should pay more attention to project publicity and information exchange with investors when launching crowd-funding projects,and pay more attention to other related crowd-funding projects to strengthen the sponsors' own qualifications and background.In addition,this paper applied four learning models,namely gradient lifting decision tree,random forest,nearest neighbor algorithm and BP neural network,which all obtained good training effects in the data set of this study,among which random forest had the best effect.The four models were compared and analyzed respectively,the advantages and disadvantages of each model in this study were summarized,and the feasibility of machine learning model application in the field of crowdfunding was verified.Finally,based on the constructed index system of influencing factors,experimental research results and the current situation of incentive crowdfunding,this paper puts forward relevant suggestions for the key aspects that fundraisers should pay attention to when launching projects and points out possible research directions for subsequent studies.