BARROS, T. S.; http://lattes.cnpq.br/7401639950436351; BARROS, Thierry Silva.
Résumé:
Geolocation data has been widely used for the
comprehension of various social phenomena. Nowadays, such
data are produced at large scale by people, especially by their
smartphones equipped with various types of sensors. However,
the capture os this kind of data, in a mobile device, can be
expensive as the collection rate increases and may consume
device resources. On the other hand, the capture with low
frequency may impair the quality and consistency of the
information collected. In this context, we conducted a
comparative of performance across different data collection
frequencies (of 15, 30, 60 and 120 seconds) to analyze the impact
on resource consumption and data quality. For this study, we used
an application developed in a previous work [1]. This app is able
to collect geolocated data from several sensors embedded in the
smartphone. The experiment had 10 volunteers, who used the
application for a period of 20 days. Afterwards, an evaluation was
performed taking into account aspects related to the device
resource consumption and aspects regarding the quality of the
data captured, in order to show the pros and cons of the different
capture frequencies and estimate a frequency best suited for
different usage scenarios. As a result, the 30-second frequency
collection had the best tradeoff between resource consumption
and information generation, and the 60-second frequency
collection was the only one that showed no better effectiveness in
any specific application compared to the other frequencies.
