Research On Personal Ized Recommendation For Crowdfunding Platform Based On Evolutionary Algor Ithms

With the gradual maturation of Internet technology,the way of raising funds has changed greatly,and the drawbacks of traditional offline financing models have gradually been exposed,for example,the lack of sufficient trust between borrowers and contributors or the lack of reasonable transparency in the financing process often leads to the failure of transactions between the two sides.There is an increasing need for an efficient and secure financing platform that meeting the needs of individuals.Crowdfunding has emerged as a novel form of financing,as a means of online financing for the general public,uses online funds from Internet users to raise funds for a particular project because of its low investment quota and threshold.Crowdfunding is able to provide opportunities for ordinary personal financing,thus be welcomed by the public.Especially with the development in recent years,the data sets of Crowdfunding platform are massive,however,platform revenue has not been matched by the growth.In other words,the platform is faced with the "information overload" problem,that is,investors can not quickly pick out platform products that can meet their personal interest preferences in a large number of products.According to the survey of the existing Crowdfunding platform,vast majority of the platform systems only provide the function of classification sorting,at this time,the Crowdfunding platform urgently needs a personalized recommendation system,which can help users to pick out the products that can meet their personal interest preferences and return needs.With the exploration in recent years,the personalized recommendation algorithms have made certain achievements.They can filter out products with high precision for users in a huge data sets,but the current personalized recommendation algorithms are mainly to capture the user's personal preferences by constructing the recommendation model or mixing different data features.Most of these algorithms evaluate the quality of the recommendation list through a single precision metric.Traditional personalized recommendation algorithm have made achievements in the book reviews network platforms or film reviews network platforms or other network platforms.But products in the Crowdfunding platform are different from books,movies and other products,for the vast majority of users in the Crowdfunding platform,investment profit is the main factor in considering whether to contribute to a product or not.Most of the traditional recommendation algorithms ignore the importance of the benefit brought by the recommendation list.At the same time,for the Crowdfunding platform,it is hoped that the recommendation list will have high diversity,which can improve the success rate of the products in the platform.Based on this,this thesis proposes a personalized recommendation algorithms for Crowdfunding platform based on evolutionary algorithms,which provides recommendation list with high profit and high diversity for users at the expense of a little precision(within the user's acceptable range).The main contributions of this thesis are described as follows:(1)In this thesis,we first propose a personalized recommendation algorithm for Crowdfunding platform based on a single-objective evolutionary algorithm,that can provide a recommendation list with high profit and high diversity for users at the expense of some accuracy.In SOEA-PRCP,two metrics,utility-accuracy and topic-diversity,are proposed to measure the quality of the recommended list in response to the need for personalized recommendations on Crowdfunding platforms.Utility-accuracy is obtained by probabilistic propagation ideas,topic-diversity is evaluated by recommendation coverage.The two sub-metrics is linearly weighted to a single-objective optimization problem by parameter weight X.Then,based on this single-objective optimization problem,a personalized recommendation algorithm based on single-objective evolutionary algorithm,SOEA-PRCP,is proposed for Crowdfunding platform.In SOEA-PRCP,in order to improve the profit of the recommendation list,an initialization strategy based on the change of product ranking is proposed.The experimental data is derived from a real Crowdfunding platform,Indiegogo.In order to verify the quality of the proposed algorithm,we run SOEA-PRCP,and several benchmark algorithms based on this datasets.Compared with the traditional recommendation algorithms,the experimental results clearly verify the validity of SOEA-PRCP and the effectiveness of the proposed strategy.(2)This thesis also presents a personalized recommendation algorithm for Crowdftunding platform based on a multi-objective evolutionary algorithm,MOEA-PRCP can provide a set of recommendation lists to meet the different preferences of the user in only one run.Through the first research work,we can find that although the performance is well in profit and diversity indicators obtained by SOEA-PRCP,which solve the personalized recommendation problem from the single-objective optimization perspective,However,the single objective optimization method can only provide a single recommendation list for users at one time,it cannot meet users' different interests and preferences,and it is inconvenient for users to set weight parameter.In order to improve users' investment experience and improve the performance of personalized recommendation service system,a personalized recommendation algorithm for Crowdfunding platform based on a multi-objective evolutionary algorithm is proposed in this thesis,which can more comprehensively improve the quality of the recommended list by providing each user with a set of good recommendation lists.Unlike SOEA-PRCP.the MOEA-PRCP maximize the two conflicting metrics,utility-accuracy and topic-diversity.Based on this multi-objective optimization problem,MOEA-PRCP is proposed.In MOEA-PRCP,in order to more comprehensively improve the quality of the recommended list,a cross strategy of parent local optimal gene reserve is proposed.Compared with SOEA-PRCP algorithm,MOEA-PRCP algorithm can provide users with multiple optimal recommendation lists to meet their diversified investment needs,and the evaluation indexes of the experimental results are significantly improved.Compared with the traditional recommendation algorithms,the experimental results also verify the effectiveness of MOEA-PRCP in the personalized recommendation on Crowdfunding platform,as well as the effectiveness of the proposed strategy.