Project recommendation using heterogeneous traits in crowdfunding

Crowdfunding has gained widespread popularity in recent years. By funding entrepreneurs with creative minds, it is gradually taking over the role of venture capitalists who provide the much needed seed capital to jump start business ventures. Despite the huge success of the crowdfunding platforms, not every project is successful in reaching its funding goal. Therefore, in this paper, we intend to answer the following question "what set of features determine a project's success?".We begin by studying the dynamics of Kickstarter, a popular reward-based crowdfunding platform, and the impact of social networks on this platform. Contrary to previous studies, our analysis is not restricted to project-based features alone; instead, we expand the features into four different categories: temporal traits, personal traits, geo-location traits, and network traits. Using a comprehensive dataset of 18K projects and 116K tweets, we provide several unique insights about these features and their effects on the success of Kickstarter projects. Based on these insights, we build a supervised learning framework to learn a model that can recommend a set of investors to Kickstarter projects. By utilizing features from the first three days of project duration alone, we show that our results are significantly better than the previous studies.