Semi-supervised learning: Predicting activities in Android environment

• Type: Article in International Conference Proceedings Book
• Date: 2012

Authors

Lopes, A (Author) Other; João Mendes-Moreira (Author) FEUP; João Gama (Author) FEP

Conference proceedings International

Title: CEUR Workshop Proceedings

Pages: 38-42

Workshop on Ubiquitous Data Mining, UDM 2012 - In Conjunction with the 20th European Conference on Artificial Intelligence, ECAI 2012

27 August 2012 through 31 August 2012

Indexing

Scopus - 2 Citations

Other information

Authenticus ID: P-00A-8QC

Abstract (EN): Predicting activities from data gathered with sensors gained importance over the years with the objective of getting a better understanding of the human body. The purpose of this paper is to show that predicting activities on an Android phone is possible. We take into consideration different classifiers, their accuracy using different approaches (hierarchical and one step classification) and limitations of the mobile itself like battery and memory usage. A semi-supervised learning approach is taken in order to compare its results against supervised learning. The objective is to discover if the application can be adapted to the user providing a better solution for this problem. The activities predicted are the most usual in everyday life: walking, running, standing idle and sitting. An android prototype, embedding the software MOA, was developed to experimentally evaluate the ideas proposed here.

Language: English

Type (Professor's evaluation): Scientific