Google has successfully utilized machine learning technology to create accurate personalized, situational app recommendation systems, having boosted the installation rate for apps at Google Play by 3.3%, according to Chi Huai-hsin, Google's chief AI researcher.
Chi, a native Taiwan talent, made the remarks when sharing his research team's AI-based app recommendation systems at a recent AI innovation boot camp activity held in Taipei.
At the moment, there are over one million apps available for users of two billion active Android-based devices, and they saw over 82 billion downloads in 2017. All these related numbers will certainly continue to expand in the future, and how to boost the app installation rate has become a very important task for Chi's research team.
Chi said that in order to provide users with better recommendation quality, Google's recommendation systems have three major principles to follow. First, recommended app contents must boast multiplicity and meet personalized needs. Second, the interfaces for all the app products must be constantly optimized with the assistance of machine learning. Third, Google recommendation systems must be applicable to all the users instead of specific ones.
In order to increase app downloads from Google Play, user preferences must be known first, and personalized recommendation systems must be established to cater to the needs of different identities and age groups, Chi said.
Then situational elements must be added to such systems to more accurately recommend apps to users. For instance, users of tablets usually prefer entertainment video apps, and smartphone users like to download apps associated with their daily lives such as living tools, emails, calendars, and notebooks. In addition, news apps usually see more downloads at daytime, while gaming apps are more preferred at night, according to Chi.