TY - GEN
T1 - AppNet
T2 - 3rd ACM SIGSOFT International Workshop on App Market Analytics, WAMA 2019, co-located with ESEC/FSE 2019
AU - Guo, Qian
AU - Wang, Haoyu
AU - Zhang, Chenwei
AU - Guo, Yao
AU - Xu, Guoai
N1 - Publisher Copyright:
Copyright © 2019 by the Association for Computing Machinery, Inc (ACM).
PY - 2019/8/27
Y1 - 2019/8/27
N2 - With the prevalence of smartphones, mobile apps have seen widespread adoption. Millions of apps in markets have made it difficult for users to find the most interesting and relevant apps. App markets such as Google Play have deployed app recommendation mechanisms in the markets, e.g., recommending a list of relevant apps when a user is browsing an app, which naturally forms a network of app recommendation relationships. In this work, we seek to shed light on the app relations from the perspective of market recommendation. We first build "AppNet", a large-scale network containing over 2 million nodes (i.e., Android apps) and more than 100 million edges (i.e., the recommendation relations), by crawling Google Play. We then investigate the "AppNet" from various perspectives. Our study suggests that AppNet shares some characteristics of human networks, i.e., a large portion of the apps (more than 69%) have no incoming edges (no apps link to them), while a small group of apps dominate the network with each having thousands of incoming edges. Besides, we also reveal that roughly 147K (7%) apps form a fully connected cluster, in which most of the apps are popular apps, while covering 97% of all the edges. The results also reveal several interesting implications to both app marketers and app developers, such as identifying fraudulent app promotion behaviors, improving the recommendation system, and enhancing the exposure of apps.
AB - With the prevalence of smartphones, mobile apps have seen widespread adoption. Millions of apps in markets have made it difficult for users to find the most interesting and relevant apps. App markets such as Google Play have deployed app recommendation mechanisms in the markets, e.g., recommending a list of relevant apps when a user is browsing an app, which naturally forms a network of app recommendation relationships. In this work, we seek to shed light on the app relations from the perspective of market recommendation. We first build "AppNet", a large-scale network containing over 2 million nodes (i.e., Android apps) and more than 100 million edges (i.e., the recommendation relations), by crawling Google Play. We then investigate the "AppNet" from various perspectives. Our study suggests that AppNet shares some characteristics of human networks, i.e., a large portion of the apps (more than 69%) have no incoming edges (no apps link to them), while a small group of apps dominate the network with each having thousands of incoming edges. Besides, we also reveal that roughly 147K (7%) apps form a fully connected cluster, in which most of the apps are popular apps, while covering 97% of all the edges. The results also reveal several interesting implications to both app marketers and app developers, such as identifying fraudulent app promotion behaviors, improving the recommendation system, and enhancing the exposure of apps.
KW - Android
KW - App Store
KW - App recommendation
KW - Google Play
UR - https://www.scopus.com/pages/publications/85076470687
U2 - 10.1145/3340496.3342757
DO - 10.1145/3340496.3342757
M3 - 会议稿件
AN - SCOPUS:85076470687
T3 - WAMA 2019 - Proceedings of the 3rd ACM SIGSOFT International Workshop on App Market Analytics, co-located with ESEC/FSE 2019
SP - 19
EP - 25
BT - WAMA 2019 - Proceedings of the 3rd ACM SIGSOFT International Workshop on App Market Analytics, co-located with ESEC/FSE 2019
A2 - Sarro, Federica
A2 - Nayebi, Maleknaz
PB - Association for Computing Machinery, Inc
Y2 - 27 August 2019
ER -