Maximizing Computing Device Runtime with Predictive Battery Life Analysis

The project proposes an end-to-end approach to training a ML model to predict normal/abnormal battery life in real-time for users of computing devices. It also identifies top contributor when user is getting ‘abnormal’ battery life. User/IT can act to improve battery life. The model is light weight so it can be executed in the background without impacting user’s productivity. The project proposes methods to label battery life data for training based on system usage, as there is no data element that provides that information. Battery life is dependent is specific to each system and is dependent on system usage, environment factors.