Data Collection and Analysis of Energy Consumption of Mobile Phones using Machine Learning Techniques

In a modern-day society there is the consensus that smartphones have a dominant role in everyday life. By just pressing a button, someone can not only get up to speed with the current events on a global scale, but also get in touch with people all over the world and find various forms of entertainment. In particular, one of the features that makes smartphones so attractive is the portability they offer, since they utilize batteries. However, batteries have a certain amount of charges in their disposal, consequently the lifespan of a device is directly correlated to its utilization, as well as its charging strategy.

The current thesis focuses on the analysis of mobile phones’ usage and the prediction of the battery’s energy drain. To begin with, for data collection the application “BatteryApp”, which periodically keeps record of the device’s usage and the battery information, was developed. The next step is the grouping of similar uses of devices through Hierarchical Clustering, which does not require an a priori selection for a specific cluster number and does not set limitations regarding the chosen distance function. After that, it was assessed based on its content in order to select the clusters with the higher information value. Lastly, the prediction of the energy drain was constructed by employing a simple linear model, two variants of linear regression, where the penalty concept is introduced (Ridge and Lasso Regression), and a non-linear model, which belongs to the Ensemble Learning category (eXtreme Gradient Boosted trees), with the parameters’ learning procedure being applied to each selected cluster individually.

Georgios Balaouras
Electrical & Computer Engineering Department
Aristotle University of Thessaloniki, Greece
October 2020

pip install numpy
pip install pandas
pip install scikit-learn
pip install scipy
pip install gower
pip install seaborn
pip install matplotlib
pip install xgboost

├── ..
 |       ├── BatteryApp: Contains the source code of BatteryApp.
 |       ├── experiment_1: all Sessions of one user.
 |       ├── experiment_2: Sessions under 30 minutes of all users with at least 20 files.
 |       ├── experiment_3: Sessions under 30 minutes of all users with at least 20 files & same total battery capacity.
 |       ├── server: Contains the scripts for hosting the server and the database.
 |       ├── preprocessing: Contains the scripts for exporting and preprocessing the data.
 |       ├── data/csvFiles: Contains the files as exported and checked from the Database.

			&
__version__	8	3.7.7	1.1.2 & 1.4.4	4.4.0

Reach out to me:

• mpalaourg's email