Session: CIE-05-01CIE Graduate Student Poster Symposium

Paper Number: 74626

Start Time: August 18, 10:00 AM

74626 - Data-Driven Recommender System for Crowdsourcing Initiatives Design 

Crowdsourcing initiatives design has been a heated topic as the applications of crowdsourcing on solving modularized problems for companies becomes more and more popular. However, the study of crowdsourcing design has been mostly focused on the theoretical side, using mathematical tools like game theory to address crowdsourcing design problems. This project offers a new perspective from the experience of real-life applications. An empirical dataset scraped from two major crowdsourcing platforms is constructed and relevant factors that are key determinants to the success of crowdsourcing initiative are extracted. The derived dataset is statistically analyzed with machine learning algorithms. After finding the dataset may be too small to be capable to train a prediction model that can fully embody the relationships hidden in the dataset, we incorporate existing theories from literature in the loop to generate synthetic dataset. Along the way, theoretical findings about crowdsourcing initiative design are examined with our empirical dataset to help researchers reflect on their work. Besides verifying existing theoretical knowledge about crowdsourcing design along building synthetic dataset, we are able to construct a recommender system with the dataset combining the empirical dataset and the synthetic dataset to give specific recommendations about crowdsourcing initiative designs once given the problem description part of factors and the expected outcome.

Presenting Author: Ziqing Li Beijing Institute of Technology

Authors: