Incremental Hierarchical Clustering of Stochastic Pattern-Based Symbolic Data

Classic data analysis techniques generally assume that variables have single values only. However, the data complexity during the age of big data has gone beyond the classic framework such that variable values probably take the form of a set of stochastic measurements instead. We refer to the above case as the stochastic pattern-based symbolic data where each measurement set is an instance of an underlying stochastic pattern. In such a case, non existing classic data analysis approaches, such as the crystal item or fuzzy region ones, could apply yet. For this reason, we put forward a novel Incremental Hierarchical Clustering algorithm for stochastic Pattern-based Symbolic Data (IHCPSD). IHCPSD is robust to overlapping and missing measurements and well adapted for incremental learning. Experiments on synthetic and application on real-life emitter parameter data have validated its effectiveness.