Three-way neighborhood based stream computing for incomplete hybrid information system.

Stream computing paradigm focuses on efficiently knowledge updating when immigration and emigration occur simultaneously. Existing stream computing for reasoning with three-way decisions aims at symbolic and complete data. In contrast, the studies on more realistic cases of numerical and symbolic types with potentially incomplete data are preliminary. To efficiently cope with the concept uncertainty, a dynamic neighborhood stream three-way decisions algorithm for incomplete hybrid information systems (DNS3WD) is proposed. Following the setting of flowing into one object and flowing out of another, we dynamically generate all neighborhood classes and decision equivalence classes in the data streaming window. Finally, we deduce the principles of variations of conditional probabilities and update three-way regions accordingly. Theoretically, the time complexity of the non-incremental three-way decisions algorithm (NIH3WD) is about N times that of DNS3WD, where N is the size of the data streaming window. In addition, extensive comparisons against NIH3WD demonstrate that DNS3WD can efficiently leverage the previously acquired knowledge. [ABSTRACT FROM AUTHOR]