Introduction

Regardless of what operating systems are being used, most users are familiar with the “loading” icon, which signals an unpleasant user wait experience. The amount of waiting time can be varied across the devices and the applications that are launched; for instance, some apps such as MSPaint, Zoom, and Fornite can take approximately 9, 12, and 16 seconds on average to launch respectively [1]. As the problem adversely affects user experience, and limited research was previously conducted in this field, we perform a study to collect user-app interaction data, analyze past behaviors to understand the user-wait events, and propose solutions to reduce such waiting times. In particular, we collect user data using Intel’s XLSDK over multiple weeks, which stands for Input, Actuator, or Logger Libraries Software Development Kit [2]. Using these data with Statistical and Machine Learning methods – namely Hidden Markov Model and Long Short-Term Memory (LSTM) model – we will be able to predict the usual wait time when a user opens an app so that we can launch the app beforehand, and the user can quickly perform their crucial tasks on the app without the long and unnecessary waiting time.