[Paper Review] Learning Mobile App Usage Routine through Learning Automata

Ramin Rahnamoun, Reza Rawassizadeh+1 more|arXiv (Cornell University)|Aug 11, 2016

Caching and Content Delivery14 references3 citations

TL;DR

This paper proposes an online learning automata (LA)-based method to predict mobile app usage routines by learning transition probabilities between app launches in real time. By dynamically updating internal states using user behavior data, the approach improves app launch prediction, reducing launch time and preserving battery life—demonstrating superior performance over two baseline methods on a real-world lifelogging dataset.

ABSTRACT

Since its conception, smart app market has grown exponentially. Success in the app market depends on many factors among which the quality of the app is a significant contributor, such as energy use. Nevertheless, smartphones, as a subset of mobile computing devices. inherit the limited power resource constraint. Therefore, there is a challenge of maintaining the resource while increasing the target app quality. This paper introduces Learning Automata (LA) as an online learning method to learn and predict the app usage routines of the users. Such prediction can leverage the app cache functionality of the operating system and thus (i) decreases app launch time and (ii) preserve battery. Our algorithm, which is an online learning approach, temporally updates and improves the internal states of itself. In particular, it learns the transition probabilities between app launching. Each App launching instance updates the transition probabilities related to that App, and this will result in improving the prediction. We benefit from a real-world lifelogging dataset and our experimental results show considerable success with respect to the two baseline methods that are used currently for smartphone app prediction approaches.

Motivation & Objective

• To address the challenge of limited smartphone battery life by improving app launch efficiency through predictive caching.
• To model and predict user app usage routines in real time using online learning techniques.
• To reduce energy consumption and enhance user experience by minimizing redundant app launches.
• To develop a dynamic, adaptive system that learns from user behavior without requiring prior knowledge or static models.
• To evaluate the effectiveness of learning automata against existing prediction methods in real-world scenarios.

Proposed method

• The system uses learning automata (LA) to model transitions between app launches as a Markov decision process.
• Each app launch event triggers an update to the transition probability matrix, refining predictions over time.
• The LA algorithm operates online, continuously adapting its internal state based on user interactions.
• The method leverages a real-world lifelogging dataset to train and validate the prediction model.
• Transition probabilities are updated incrementally using reinforcement learning principles to improve accuracy.
• The approach integrates with the OS's app caching mechanism to pre-load frequently used apps.

Experimental results

Research questions

• RQ1 Can learning automata effectively model and predict user mobile app usage routines in real time?
• RQ2 How does the online learning approach compare to existing static or batch-based prediction methods in terms of accuracy and efficiency?
• RQ3 To what extent can the predicted app usage patterns reduce app launch time and improve battery life?
• RQ4 How do dynamic updates to transition probabilities enhance prediction performance over time?
• RQ5 What is the impact of using real-world user behavior data on the robustness of the prediction model?

Key findings

• The proposed learning automata approach achieved significantly higher prediction accuracy compared to two baseline methods on the real-world lifelogging dataset.
• The system reduced app launch time by leveraging predictive caching based on learned usage patterns.
• Battery life was preserved by minimizing unnecessary app reloads through accurate prediction of user routines.
• The online learning mechanism enabled continuous improvement of prediction performance as user behavior evolved.
• The method demonstrated robustness and adaptability in real-world conditions, outperforming static prediction models.
• The results confirmed that learning automata are effective for modeling complex, dynamic user behavior in mobile environments.