New Trends in Knowledge Driven Data Mining a position paper

from classification to pattern mining, reached considerable levels of efficiency, and their extension to deal with more demanding data, such as data streams and big data, show their incontestable quality and adequacy to the problem. Despite their efficiency, their effectiveness on identifying useful information is somehow impaired, not allowing for making use of existing domain knowledge to focus the discovery. The use of this knowledge can bring significant benefits to data mining applications, by resulting in simpler and more interesting and usable models. However, most of existing approaches are concerned with being able to mine specific domains, and therefore are not easily reusable, instead of building general algorithms that are able to incorporate domain knowledge, independently of the domain. In our opinion, this requires a drift in the focus of the research in data mining, and we argue this change should be from domain-driven to knowledge-driven data mining, aiming for a stronger emphasis on the exploration of existing domain knowledge for guiding existing algorithms.