Frontiers in Health Informatics
Volume:4 Issue: 2, 2015

The past few years have shown an increase in the amount of data that are being generated in all fields. This increase is due to human needs to all aspects from business to science. One of these fields is medicine with uncertain data. For these amount of data there exists computational methods and tools for analysis and it requires medical informatics researchers and specialists to choose the most appropriate method to cope with these data. This paper presents one of the common data mining tasks, association rule mining, when considering particular continuous time series measurements. The most important goal of this study is dealing with and processing time series data for the purpose of extracting linguistic rules. Also, comparing the results of this method with the results of classical statistical analysis is the second objective. A three-part analytical pipeline is used in this study as a new mining approach, consisting of data preparation, data transformation, and data analysis to elicit knowledge about the interaction of the parameters measured and the effect of different drugs on the rats. Weka data mining toolkit has been employed for extracting knowledge in the form of association rules and the important rules identified by this toolkit are interpreted for their medical significance. Discretizing time series data and linguistic representation of rules have been demonstrated. This study exhibits this new approach by analysis of the data of the cardiovascular responses to specific medications in rats that have been recorded by lab chart software in the laboratory. Once the time series was discretized, simple association rule mining methods are used to extract rules. The association rules that are discovered can provide the domain expert with new knowledge about the possible effects of injected rats or to help to improve annotation consistency.