Machine Learning Methods for Smartphone Application Prediction

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

124 Downloads (Pure)


The growing number of smartphone apps has led to
a tremendous increase in the number of apps installed on users’
smartphones. This not only provides hard problems for users
to find the one they need, but also slows down the smartphone
and increases the operating load if running at the same time.
A precise app prediction function is desirable to improve user
experience by recommending or even automatically starting the
apps. As it is manifested as a new problem arising in recent
years, there is very limited work available, and most of the
existing studies have used inadequate datasets (limited number
of participants or applications) to build models. In this paper, a
large number of real smartphone usage records collected from
573 anonymous users are utilised. Four popular machine learning
methods, namely K-Nearest Neighbour, Support Vector Machine,
AdaBoost and Decision Tree, are initially employed to build
the models for app prediction and then their performances are
thoroughly compared via a number of experiments. Results show
that the bagging decision tree performs the best while AdaBoost
also produced a high level of accuracy. However, a support vector
machine does not appear to be suitable for this application.
Original languageEnglish
Title of host publication2022 IEEE International Symposium on Industrial Electronics
Publication statusAccepted/In press - Apr 2022


Dive into the research topics of 'Machine Learning Methods for Smartphone Application Prediction'. Together they form a unique fingerprint.

Cite this