Two-Stage Data Mining for Big Statistical Micro Data

We apply a two-stage data mining strategy to handle and analyze big statistical micro data sets. The first stage consists of smart aggregation of such micro data, and the second data continues to analyze and visualize the smartly aggregated data, further. The smart aggregation here requires three steps. One is to decide and to create the appropriate aggregates themselves, called also ‘concepts.’ Second, the characteristics for the concepts need to be implemented. The Symbolic Data Analysis (SDA) approach offers good tools for this. In this paper, our symbolic variables are of the two kinds: frequencies of categorical variables, and intervals of continuous variables. The third step is to operationalize the first two steps by creating the new data set. This operation is performed by the SYR software for SDA from Syrokko Company. We present our methodology with empirical data from the fifth round of the European Social Survey.