Design of A Universal User Model for Dynamic Crowd Preference Sensing and Decision-Making Behavior Analysis

Abstract

Sharing economy becomes an emerging issue in urban life. It is not a new phenomenon but an assembling of existing techniques to meet specific demands of users. It also points out a better way to implicitly collect users' contexts and to understand users than the conventional one that requires much user involvement (e.g., tedious inputs). A universal model, for this purpose, that supports dynamic analysis and mining of user-generated content (or contexts) is designed in this paper. Two major factors, sensing and analysis of crowd preference and their decision-making behavior, are especially targeted. This model formulates the given scenario that comprehensively illustrates the possible actors and correlated actions among them with a set of rules to enhance the machine learning results. This model outlines a detail process on pre-/post-process of the data, and indicates the core techniques for user modeling. The raw data collected from on-service website, i.e., Airbnb, are utilized for the preliminary examination of our proposal. We especially look at internal factors (e.g., nationality, gender, and age) and external factors (e.g., device, social media, and time) that reflect implicitly the difference on crowd's preference and behavior. Results after statistics-based machine learning reveal that the relation among users' internal and external factors share high similarity with their behavior patterns, and can be applied, considering particular features, for service provision to a specific type of crowds.

Keywords

• User understanding
• crowd preference
• human behavior
• statistical analysis
• human-centric computing
• sharing economy