Understanding Smartwatch Battery Utilization in the Wild.

Area of Science:

• Wearable Technology
• Computer Science
• Data Science

Background:

• Smartwatch battery limitations hinder consumer adoption.
• Limited research exists on real-world smartwatch battery usage across diverse devices.
• Understanding battery consumption patterns is crucial for improving user experience.

Purpose of the Study:

• To analyze real-world smartwatch battery utilization patterns.
• To develop an accurate model for predicting high battery discharge events.
• To introduce a method for assessing smartwatch battery quality degradation over time.

Main Methods:

• Utilized a dataset from 832 real-world smartwatch users.
• Employed clustering to identify battery utilization patterns.
• Developed a transparent convolutional neural network (CNN) model for binary classification (low/high consumption).
• Introduced a longitudinal indexing method to track battery quality changes.

Main Results:

• The CNN model achieved 85.3% accuracy in predicting high battery discharge events.
• The model outperforms existing machine learning algorithms in battery discharge prediction.
• The model allows for the extraction of information from learned filters, offering unique insights.
• The indexing method quantifies battery quality changes over time.

Conclusions:

• The developed deep learning model accurately predicts high battery discharge events.
• The findings provide actionable insights for improving smartwatch battery performance.
• This research assists manufacturers, developers, and users in optimizing smartwatch battery life.