Chargetime Predictor: A Data-Driven Approach for Estimating Smartphone Charging Session Duration

Abstract

Overcharging smartphone batteries is one of fundamental causes of battery degradation. As it can significantly impact their lifespan by the increased battery temperature it causes. Predicting the target time for unplugging the smartphone from the charger has become a hot topic in both academia and industry. In this paper, we propose a new machine learning approach using XGBoost to predict the target unplugging time, the event upon which the user will unplug the phone from the charger, based on the user's charging routine. We integrate our proposed approach into a commercial product that charges the battery at intervals based on the predicted unplugging time. This technique can almost double the average battery lifespan cycle from two to four and a half years in realworld testing. Our proposed technique can help users extend their battery life by avoiding overcharging by predicting the end time of the charging session. Our end-to-end system is introduced in this work, emphasising the prediction machine learning model. To train our model, we used a real-world dataset with different charging routines, achieving an accuracy of 0.94 and an F1-score of 0.91 in real-world testing, not limited to simulation testing. Additionally, our solution provides a reasonably accurate prediction for any charging session, unlike previous work, which focused solely on predicting the longest session.