Event Correlation with Undefined Data
N. I. Pikuleva, A. Sh. Khafizova, D. A. Gashigullin · 2021
Data mining techniques solve a numerous number of problems. One of the main challenges of these techniques is to make large data processing efficient. Researches provide enhancements for the existing algorithms using modern technologies such as Hadoop and systems build on the top of it. Growing of the data is not the only challenge we have to deal with. A lot of data comes with uncertainties, which make data processing not trivial. We have to take into consideration that uncertainty may have different forms. Examples of uncertainties are the data emitted from sensors, moving objects, events with non-accurate spatio-temporal properties, etc. This work focuses on finding correlations among the events which occurred within uncertain time and uncertain location. Finding correlations is a good field for applying clustering algorithms. Here we consider that clustering will process data in both spatial and temporal dimensions at a time. To make the algorithm efficient we implemented it using a well-established technique, MapReduce, which allows process large data sets in distributed manner. Thereby, in our work we developed framework which allows to find correlations among the uncertain events within large data sets efficiently.