Research On Personalized Recommendation Algorithm In Crowdfunding Platform

The maturation of Web 2.0 and the development of Internet makes it easier for work to be carried out from offline to online, thus appears the concept of Internet group collaboration. Crowdsourcing is a typical type of Internet group collaboration. In the business model of crowdsourcing, crowdfunding has received a lot of attention from ordinary users. As a public financing activity, crowdfunding is welcomed by people due to its low invest amount and threshold. With the help of public channels and funds, crowdfunding can provide certain material supports to projects. Crowdfunding industry has experienced rapid development since the establishment of the first crowdfunding platform, ArtistShare. According to the 2015CF Crowdfunding Industry Report released by Massolution, the total amount raised in 2015 will reach $34 billion 400 million.With the expansion of the scale of the crowdfunding platform, it is not only increasingly difficult for the platform to maintain the list of products, users are also difficult to completely browse and rapidly find products they are interested in. According to the survey on existing crowdfunding platforms, they provide only a sort capability, users cannot filter products, neither can they get recommendation service from the system. In the initial stage of crowdfunding platforms, the existing platform architecture can meet the demand of users because the amount of users and products is small. In the era of information explosion, users not only expect to be able to quickly and effectively access to information, but also hope to get intelligent recommendation from system. E-commerce and other similar platforms discovered this problem. They help users to filter and recommend information automatically with the help of a variety of personalized recommendation algorithms. They reduce the cost of users’and enhance users’satisfaction. Currently, crowdfunding platform cannot provide these services. As a result, it is necessary that personalized recommendation system should be an indispensable function in crowdfunding platform.Based on the current research state of personalized recommendation system, we try to explore the feasibility and results of the application of personalized recommendation algorithm in crowdfunding platform. Considering there is rarely literature research concerning on this area, we make use of literature research method to summarize the origin and development of personalized recommendation system, compare the advantages and disadvantages of recommendation algorithms, and choose the most suitable recommendation method according to the features of users and projects.The main contributions of this paper are confined as below:(1) We collected some typical crowdfunding platforms, made a comprehensive analysis of their structures and functions. In order to build models and do some calculations for personalized recommendation, we extracted and screened characteristics of users’and projects’of crowdfunding platforms.(2) By applying personalized recommendation algorithm to crowdfunding platforms, we made some algorithm improvements according to the characteristics of crowdfunding platforms. At last, based on actual user data collected, we verified the performance of improved algorithm in many aspects.This paper still exists limitations because of the restrictions of the lack of human and resource. Using the user data collected before, we can only carry out simple experiments to verify the performance of the algorithm. As a result, we are not able to provide further suggestions for problems may arise if the algorithm be widely used in various platforms. This paper will make further research on web content mining, combined collaborative filtering and other aspects. We will also explore a better way to improve the efficiency and reduce the time cost at the same time.