Leveraging Mobile App Classification and User Context Information for Improving Recommendation Systems

Abstract

Mobile apps play a significant role in current online environments where there is an overwhelming supply of information. Although mobile apps are part of our daily routine, searching and finding mobile apps is becoming a nontrivial task due to the current volume, velocity and variety of information. Therefore, app recommender systems provide users’ desired apps based on their preferences. However, current recommender systems and their underlying techniques are limited in effectively leveraging app classification schemes and context information. In this thesis, I attempt to address this gap by proposing a text analytics framework for mobile app recommendation by leveraging an app classification scheme that incorporates the needs of users as well as the complexity of the user-item-context information in mobile app usage pattern. In this recommendation framework, I adopt and empirically test an app classification scheme based on textual information about mobile apps using data from Google Play store. In addition, I demonstrate how context information such as user social media status can be matched with app classification categories using tree-based and rule-based prediction algorithms. Methodology wise, my research attempts to show the feasibility of textual data analysis in profiling apps based on app descriptions and other structured attributes, as well as explore mechanisms for matching user preferences and context information with app usage categories. Practically, the proposed text analytics framework can allow app developers reach a wider usage base through better understanding of user motivation and context information.