Distributed Data Mining

Abstract

The continuous developments in information and communication technology have recently led to the appearance of distributed computing environments, which comprise several, and different sources of large volumes of data and several computing units. The most prominent example of a distributed environment is the Internet, where increasingly more databases and data streams appear that deal with several areas, such as meteorology, oceanography, economy and others. In addition the Internet constitutes the communication medium for geographically distributed information systems, as for example the earth observing system of NASA (eos. gsfc.nasa.gov). Other examples of distributed environments that have been developed in the last few years are sensor networks for process monitoring and grids where a large number of computing and storage units are interconnected over a high-speed network. The application of the classical knowledge discovery process in distributed environments requires the collection of distributed data in a data warehouse for central processing. However, this is usually either ineffective or infeasible for the following reasons: (1) Storage cost. It is obvious that the requirements of a central storage system are enormous. A classical example concerns data from the astronomy science, and especially images from earth and space telescopes. The size of such databases is reaching the scale of exabytes (1018 bytes) and is increasing at a high pace. The central storage of the data of all telescopes of the planet would require a huge data warehouse of enormous cost. (2) Communication cost. The transfer of huge data volumes over network might take extremely much time and also require an unbearable financial cost. Even a small volume of data might create problems in wireless network environments with limited bandwidth. Note also that communication may be a continuous overhead, as distributed databases are not always constant and unchangeable. On the contrary, it is common to have databases that are frequently updated with new data or data streams that constantly record information (e.g remote sensing, sports statistics, etc.). (3) Computational cost. The computational cost of mining a central data warehouse is much bigger than the sum of the cost of analyzing smaller parts of the data that could also be done in parallel. In a grid, for example, it is easier to gather the data at a central location. However, a distributed mining approach would make a better exploitation of the available resources. (4) Private and sensitive data. There are many popular data mining applications that deal with sensitive data, such as people’s medical and financial records. The central collection of such data is not desirable as it puts their privacy into risk. In certain cases (e.g. banking, telecommunication) the data might belong to different, perhaps competing, organizations that want to exchange knowledge without the exchange of raw private data. This article is concerned with Distributed Data Mining algorithms, methods and systems that deal with the above issues in order to discover knowledge from distributed data in an effective and efficient way.