Paper Details

ABSTRACT

More and more data mining applications are running on mobile devices such as ‘Tablet PCs’, smart phones and Personal Digital Assistants (PDAs). The ability to make phone calls and send SMS messages nowadays seems to be merely an additional feature rather than the core functionality of a smart phone. Smartphone’s offer a wide variety of sensors such as cameras and gyroscope as well as network technologies such as Bluetooth, and Wi-Fi with which a variety of different data can be generated, received and recorded. Furthermore smartphones are computationally able to perform data analysis tasks on these received, or sensed data such as data mining. Many data mining technologies for smartphones are tailored for data streams due to the fact that sensed data is usually received and generated in real-time, and due to the fact that limited storage capacity on mobile devices requires that the data is analyzed and mined on the fly while it is being generated or received. For example, the Open Mobile Miner (OMM) tool allows the implementation of data mining algorithms for data streams that can be run on smartphones. However, to the best of our knowledge, all existing data mining systems for mobile devices either facilitate data mining on a single node or follow a centralized approach where data mining results are communicated back to a server which makes decisions based on the submitted results.