Applying NoSQL Databases for Operationalizing Clinical Data Mining Models

Abstract

Access to data mining models built in clinical data systems is limited to relatively small groups of researches, while they should be available in real-time to clinicians in order to deliver the results at the point where it is most useful. At the same time, complexity of data processing grows as volume of available data exponentially rises and includes unstructured data. Clinical decision support systems based on relational and multidimensional technology lack capabilities of processing all available data because of its volume and format. On the other hand, NoSQL repositories offer great flexibility and speed in terms of data processing, but requires programming skills. A proposed solution presented in this paper is to combine both of the technologies in a single analytical system. Dual view of the data gathered in the repository allows to use data-mining tools, while Big Data technology delivers necessary data. Key-value style of querying a database enables efficient retrieval of input data for analytical models. Online loading processes guarantee that data is available for analysis immediately after it is produced either by physicians or medical equipment. Finally, this architecture can be successfully moved to the cloud. © Springer International Publishing Switzerland 2014.