Elsevier

Decision Support Systems

Volume 91, November 2016, Pages 67-76
Decision Support Systems

The determinants of crowdfunding success: A semantic text analytics approach

https://doi.org/10.1016/j.dss.2016.08.001Get rights and content

Highlights

  • A text analytics methodology is proposed for analyzing and predicting crowdfunding success.

  • Domain-constraint LDA is developed to extract semantic features from texts.

  • State-of-the-art feature selection and data mining methods are explored.

  • Our research contributes to advance the computational method for analyzing crowdfunding success.

Abstract

In the era of the Social Web, crowdfunding has become an increasingly more important channel for entrepreneurs to raise funds from the crowd to support their startup projects. Previous studies examined various factors such as project goals, project durations, and categories of projects that might influence the outcomes of the fund raising campaigns. However, textual information of projects has rarely been studied for analyzing crowdfunding successes. The main contribution of our research work is the design of a novel text analytics-based framework that can extract latent semantics from the textual descriptions of projects to predict the fund raising outcomes of these projects. More specifically, we develop the Domain-Constraint Latent Dirichlet Allocation (DC-LDA) topic model for effective extraction of topical features from texts. Based on two real-world crowdfunding datasets, our experimental results reveal that the proposed framework outperforms a classical LDA-based method in predicting fund raising success by an average of 11% in terms of F1 score. The managerial implication of our research is that entrepreneurs can apply the proposed methodology to identify the most influential topical features embedded in project descriptions, and hence to better promote their projects and improving the chance of raising sufficient funds for their projects.

Introduction

In the era of the Social Web, crowdfunding has become an increasingly more important way for entrepreneurs or small enterprises to raise the essential capitals from the crowd to support their projects or businesses. Crowdfunding websites such as Kickstarter and IndieGoGo behave as online intermediary agents that allow projects founders to quickly reach a large number of individual investors with minimal costs. It is believed that crowdfunding helps converting ordinary customers to business investors [1]. Not only does the idea of crowdfunding emerge in developed countries but it also becomes very popular in developing countries such as China. According to the statistics revealed at the crowdfunding website named Zhongchou1 in China, there were 15,073 projects supported by 802,308 backers who contributed a total of $171,753,514 (RMB) by 2014. Dreamore, Demohour, and Zhongchou are among the most popular crowdfunding websites in China. However, the percentage of crowdfunding projects that can reach their original fund raising goals is relatively small among all crowdfunding websites. For example, only 44% of the projects can reach their initial fund raising goals at Kickstarter [2]. For most crowdfunding websites, if a project cannot raise sufficient fund in time with respect to the original funding goal, the project will be marked as a failure and its fund raising campaign will stop. This becomes the main hindrance of leveraging crowdfunding to acquire venture capitals for supporting a variety of innovative business ideas. In this paper, we focus on examining the influential features, particularly textual features which may affect fund raising successes of crowdfunding projects. For brevity, crowdfunding success refers to the fund raising success of a project for the rest of this paper.

Previous research has examined the dynamics of founders and backers on crowdfunding platforms [1], [3], [4], [5], [6], [7]. However, these studies mainly examined crowdfunding platforms of the developed countries such as Kickstarter and IndieGoGo that used English as the main communication language. Few studies have investigated into the crowdfunding dynamics in developing countries such as China. Moreover, though some studies explored the impact of numerical features (e.g., funding amount, duration of project, etc.) on crowdfunding success [1], [8], [9], none of the previous study examined the influence of topical features (i.e., latent semantics) mined from textual descriptions of projects on fund raising success. Our research work is just able to fill the aforementioned research gaps.

For crowdfunding, project founders raise funds by describing their projects and offering rewards to investors (i.e., backers) via crowdfunding websites. Accordingly, the descriptions of projects and rewards can be leveraged to analyze and predict project success. For instance, we successfully extracted a set of semantically related terms (i.e., a topic) such as {protection, rubbish, environment, ecology, …} by applying a topic modeling method [10] to the project descriptions published at Dreamore. This topical feature turns out to be effective to predict fund raising success because the trend in China and the rest of the world is toward environmental protection, and individuals generally like to support “greeny” projects. For topic modeling, each topic captures a semantically coherent “concept” about the real-world (e.g., environmental protection) [11], [12]. One novelty of our research is to combine topical features with common numerical features to enhance the prediction of crowdfunding success. A topic-based text analytics method is different from the traditional keyword-based approach in that a topic consists of a set of semantically coherent words, whereas the keyword-based method assumes the independence among words [11], [13]. For instance, given the keyword “apple”, it may refer to projects of Apple Inc. that are generally supported by the crowd, or projects related to “apple fruit” that are not necessarily supported by the crowd. In fact, previous research shows that keyword-based method is not as effective as topic-based method in text analytics [11], [13], [14].

In sum, the main contributions of our research work are threefold. First, we design a novel text analytics framework for analyzing and predicting fund raising successes of crowdfunding projects. Second, we develop the domain-constraint LDA (DC-LDA) topic model for more effective mining of topical features (i.e., latent semantics) from textual descriptions of projects. Finally, we performed an empirical analysis to identify the discriminatory features that influence fund raising successes of projects based on real-world crowdfunding platforms. To the best of our knowledge, this is the first successful research of applying a topic modeling method to extract topical features from project descriptions to analyze and predict crowdfunding project success. The managerial implication of our research is that entrepreneurs or small businesses can apply the proposed framework to identify the most influential textual features that affect fund raising outcomes, and hence they can publicize and promote their startup projects by using these features to improve the chance of fund raising success.

The rest of the paper is organized as follows. Section 2 summarizes previous studies related to crowdfunding, and compares our work with the previous studies. Then, the proposed text analytics-based methodology for analyzing and predicting crowdfunding success is highlighted in Section 3. Section 4 illustrates the computational details of the proposed methodology. Discussions of the experimental procedures and the empirical results are given in Section 5. Finally, we offer concluding remarks and pinpoint future directions of our research work.

Section snippets

An overview of crowdfunding

While crowdsourcing aims to help people perform various tasks by leveraging the “crowd” [15], crowdfunding refers to the completion of a specific type of task, that is, fund raising by using the crowd. Crowdfunding is “an open call, essentially through the Internet, for the provision of financial resources either in form of donation or in exchange for some form of reward and/or voting rights in order to support initiatives for specific purposes” [16]. Crowdfunding consists of three important

A text analytics framework for crowdfunding analysis

Though previous studies found that the language used in project descriptions might influence fund raising success [9], only shallow linguistic features (e.g., number of words, spelling errors, etc.) were examined. For the proposed text analytics framework, we advocate a topic modeling method to extract topical features (i.e., latent semantics) from project descriptions and reward descriptions to predict crowdfunding success. In particular, we design a new DC-LDA topic model by extending the

The classical LDA model

Topic modeling methods have been widely used in text mining tasks. Latent Dirichlet Allocation (LDA) is one of the topic modeling methods that introduces a latent variable (i.e., a topic) between the traditional constructs of documents and words to better describe the generation of each document of a textual corpus [10]. The topics represent the latent semantic concepts embedded in documents [11], [13]. More specifically, the LDA model utilizes two Dirichlet distributions, namely topic-word

The dataset

Our crowdfunding dataset was retrieved from two popular crowdfunding websites in China, namely Dreamore2 and Zhongchou by using our dedicated crawlers developed with Python. Moreover, we accessed to the Sina online news forum3 which is one of the most popular portals in China. All the online news articles referring to “crowdfunding” were crawled for the construction of the semantic network. At the end, 23,113 online news articles were collected

Conclusions and future work

In the era of the social Web, crowdfunding has emerged to be one of the most important funding sources for entrepreneurs or small enterprises. While previous studies have identified some numerical features for predicting crowdfunding success, little research is conducted to explore the topical features (e.g., the latent semantics embedded in project or reward descriptions) for crowdfunding analysis. Our research work is just able to fill the aforementioned research gap. The main contributions

Acknowledgements

This work was supported in part by the 973 Project (No. 2012CB316205), Humanities and Social Sciences Foundation of the Ministry of Education (No. 14YJA630075), Beijing Nova Program (No. Z131101000413058) and the Fundamental Research Funds for the Central Universities, and the Research Funds of Renmin University of China (No. 15XNLQ08). Lau's work was supported by grants from the Research Grants Council of the Hong Kong Special Administrative Region, China (Projects: CityU 11502115 and CityU

Ms. Yuan is a Ph. D. student at Department of Information Systems, City University of Hong Kong. She got her bachelor degree in Information Systems and master degree in Management Science and Engineering at School of Information, Renmin University of China. Her research interests include web mining, business intelligence and decision support systems.

References (49)

  • A. Ordanini et al.

    Crowdfunding: transforming customers into investors through innovative service platforms

    Journal of Service Management

    (2011)
  • V. Etter et al.

    Launch hard or go home: predicting the success of Kickstarter campaigns

  • E.M. Gerber et al.

    Crowdfunding: motivations and deterrents for participation

    ACM Transactions on Computer-Human Interaction

    (2013)
  • M. Muller et al.

    Crowdfunding inside the enterprise: employee-initiatives for innovation and collaboration

  • G. Burtch et al.

    An empirical examination of peer referrals in online crowdfunding

  • C.T. Lu et al.

    Inferring the impacts of social media on crowdfunding

  • M.D. Greenberg et al.

    Crowdfunding support tools: predicting success & failure

  • T. Mitra et al.

    The language that gets people to give: phrases that predict success on Kickstarter

  • D.M. Blei et al.

    Latent Dirichlet allocation

    The Journal of Machine Learning Research

    (2003)
  • R.Y.K. Lau et al.

    A probabilistic generative model for mining cybercriminal networks from online social media

    IEEE Computational Intelligence Magazine

    (2014)
  • I. Titov et al.

    Modeling online reviews with multi-grain topic models

  • G. Xu et al.

    Named entity mining from click-through data using weakly supervised latent Dirichlet allocation

  • F. Kleemann et al.

    Un (der) paid innovators: the commercial utilization of consumer work through crowdsourcing

    Science, Technology & Innovation Studies

    (2008)
  • P. Belleflamme et al.

    Crowdfunding: tapping the right crowd

    Journal of Business Venturing

    (2013)
  • Cited by (176)

    • The impact of the reward scheme design on crowdfunding performance

      2023, Technological Forecasting and Social Change
    View all citing articles on Scopus

    Ms. Yuan is a Ph. D. student at Department of Information Systems, City University of Hong Kong. She got her bachelor degree in Information Systems and master degree in Management Science and Engineering at School of Information, Renmin University of China. Her research interests include web mining, business intelligence and decision support systems.

    Dr. Raymond Y.K. Lau is an associate professor at Department of Information Systems, City University of Hong Kong. His research interests include Big Data Stream Analytics, Text Mining, and e-Commerce. He has published over 150 research papers in international journals and conferences, such as MIS Quarterly, INFORMS Journal on Computing, Decision Support Systems, ACM Transactions on Information Systems, and IEEE Transactions on Knowledge and Data Engineering. He is a senior member of the IEEE and the ACM, respectively.

    Dr. Xu is an associate professor at School of Information, Renmin University of China. He is a research fellow at Department of Information Systems, City University of Hong Kong. He got his bachelor and master degree in Mathematics at Xi'an Jiaotong University and doctor degree in Management Science at Chinese Academy of Sciences. His research interests include web mining, business intelligence and decision support systems. He has published over 70 research papers in international journals and conferences, such as Decision Support Systems, European Journal of Operational Research, Fuzzy Sets and Systems, IEEE Trans. Systems, Man and Cybernetics, and International Journal of Production Economics.

    View full text