Paper Type

Complete

Paper Number

1262

Description

On home-sharing platforms like Airbnb, the user-generated data provided by hosts and guests are valuable for user trustworthiness prediction (UTP). They convey not only personal information associated with individual users (e.g., age and gender), but also other social cues (e.g., host-guest homophily). However, user data have an intrinsic property of heterogeneity, which contains various types of entities and connections. Additionally, previous research in UTP primarily focused on users’ personal features, leaving the social features largely unexploited. In this paper, we propose a novel heterogeneous graph framework for UTP. Particularly, we build a Heterogeneous Graph Attention network (HGAT) on a Heterogeneous Information Graph (HIG). The HIG can integrate heterogeneous information from users and capture their interconnections, whilst the HGAT selectively aggregates this information for UTP. Experiments with a real-world Airbnb dataset showed that our method performed better than other cutting-edge methods, demonstrating an effective usage of our framework for UTP.

Comments

Crowd

Share

COinS
 
Jul 2nd, 12:00 AM

Homophily Graph Networks for Trustworthiness Prediction on Airbnb

On home-sharing platforms like Airbnb, the user-generated data provided by hosts and guests are valuable for user trustworthiness prediction (UTP). They convey not only personal information associated with individual users (e.g., age and gender), but also other social cues (e.g., host-guest homophily). However, user data have an intrinsic property of heterogeneity, which contains various types of entities and connections. Additionally, previous research in UTP primarily focused on users’ personal features, leaving the social features largely unexploited. In this paper, we propose a novel heterogeneous graph framework for UTP. Particularly, we build a Heterogeneous Graph Attention network (HGAT) on a Heterogeneous Information Graph (HIG). The HIG can integrate heterogeneous information from users and capture their interconnections, whilst the HGAT selectively aggregates this information for UTP. Experiments with a real-world Airbnb dataset showed that our method performed better than other cutting-edge methods, demonstrating an effective usage of our framework for UTP.

When commenting on articles, please be friendly, welcoming, respectful and abide by the AIS eLibrary Discussion Thread Code of Conduct posted here.