On Formalism of Continuous Knowledge Discovery and Temporal Granularity

Abstract

The accessibility and abundance of data today makes knowledge discovery a matter of considerable importance and necessity. The process to discover continuously knowledge in evolving business domain is a challenge issue. A continuous knowledge discovery process is introduced for inducing the local first-order rules and global evolutional rules, to trace dynamic evolution patterns firstly. The definitions of main notions (event, sequence pattern, temporal rule) are proposed in a formal way, based on first-order linear temporal logic and temporal granularity. The measures of support and confidence about ranged degree of truth of a formula are established. The formalism defines the valuation on a linear state structure with time granules. By defining transition operation between temporal types, it is proved that only the independent information for unspanned-granule may be transferred without loss among different granularities. Otherwise, an aggregation mechanism was proposed to state sequence.