Article Preview
Top1. Introduction
Crowdfunding allows founders of profit, cultural and artistic ventures to raise money by acquiring small amounts of funding from many different small organizations and people. Crowdfunding is especially lucrative for young entrepreneurs aiming to materialize their novel ideas or set-up start-ups that are unable to acquire funding either because of lack of platform or lack of interest of Venture Capitalists (because of inexperience and lack of trust). Crowdfunding provides another but an unconventional method to accumulate funding, in contrast to the age-old methods of relying on government banks for loans or giving equity in the start-up. The projects open for crowdfunding belong to varied domains and categories such as theatre, comics, art, film, and music. Crowdfunding websites like Kickstarter (https://www.indiegogo.com) attract a large number of people to invest and in return get rewards or equity (Agrawal et al., 2015). This not only helps the entrepreneurs in raising money but also helps them to market their products. All this happens online without the involvement of any middlemen, thus distinguishing it even more from other traditional platforms. Here, the investors become a part of the journey of an entrepreneur or inventor.
Crowdfunding is preferred by those who do not want to indulge in traditional methods to fund their projects or do not wish to involve the middlemen (Gerber et al., 2012). However, out of thousands of projects, it is very difficult for an investor to determine which project will be successful. Hence, it becomes imperative to identify the most influential parameters on which a project’s outcome depends and subsequently determine or predict the success/failure of a project (Cordovaa et al., 2015). This work utilizes machine learning algorithms using predictive modeling techniques to predict the likely outcome of a crowdfunding venture based on whether it acquires the goal amount it needed initially. Two datasets that belong to two different types of popular crowdfunding websites have been utilized in the work. One of the websites is Kickstarter that is a reward-based crowdfunding platform while the other set of data is from the Indiegogo website, an equity-based crowdfunding platform. The collected data has been stored in multiple fragments, i.e., shards in a distributed database, MongoDB (n.d.). Distribution of data helps in reducing the load on a single server.
The stored data has firstly been visualized in order to find important factors that influence the fate of projects. Subsequently, machine learning algorithms use these crucial parameters/attributes of the data according to the type of crowdfunding. Machine learning algorithms prevent us from hard-coding old-school algorithms to reach our results because of their ability to automatically make the programs learn and adapt to the increasing dataset.(Thrun & Pratt, 2012). Machine learning techniques are highly useful in the current era of big data since they help us in quickly develop models that can aid us in analyzing huge or convoluted datasets and delivering quick and precise results even on a very large scale. Most of the industries, such as healthcare, marketing, sales, transportation, financial services, etc, working upon large amounts of data have adopted machine learning technology. The current work predicts the outcome of a crowdfunding project using the machine learning algorithms, including the most commonly used ones like Naïve Bayes, Support Vector Machine (SVM) (Ng & Jordan, 2002) and Random Forest to train and test the data for prediction of success of any given project.
The remaining portion of the paper is organized in the following manner: Section II presents the background study of the domain of crowdfunding and work done in this field. We discuss the data analysis methodology applied in Section III. Results of visualization and outcome prediction are put forth in Section IV. Lastly, we conclude the study.