A Multi-Objective Differential Evolutionary Method for Constrained Crowd Judgment Analysis

Abstract

Crowdsourcing has already been shown to be a promising tool in solving many real-life problems in time and cost-effective way. For example, in city planning, to install some specific facilities it is required to acquire knowledge about various factors like demand, demographic information, suitability of the resources in that area, etc. However, obtaining this information is a tedious and time-consuming job. Now-a-days, this process can be accelerated by utilizing the enormous power of crowd while outsourcing it to the general people. Basically, seeking opinions from multiple non-experts instead of a single expert can be advantageous in terms of time, cost and accuracy. Although, in most of the crowdsourcing models, the questions posted to crowd consist of a single component. Interestingly, in many real-life applications like city planning, the questions can have multiple components. To exemplify, the posted question can be seeking opinions about 2D coordinates of k best possible locations (i.e., k components) to install k facilities. Moreover, there exist some constraints which are needed to be satisfied by the crowd while providing their opinions. Thus, it introduces a new kind of judgment analysis problem recently termed as `Constrained Judgment Analysis'. Most of the state-of-the-art judgment analysis problems deal with the question without multiple components and constraints as well. In this article, we address this emerging problem and propose a multi-objective differential evolution method to obtain better decision guided by the crowd. The effectiveness of the proposed method is demonstrated by applying it over two real-life crowd opinion datasets.

Keywords

• Crowdsourcing
• judgment analysis
• multi-objective optimization