Timestamp interestingness measure for time series numerical association rule mining

Nowadays, Numerical Association Rule Mining methods, where most of them are based on population-based bio-inspired algorithms, represent a fruitful direction for extracting knowledge from transaction databases. Recently, Numerical Association Rule Mining methods have been extended to mine time-series datasets. However, these modified approaches introduce both opportunities and challenges for time-series mining, allowing users to uncover interesting temporal segments within the data. One of the biggest difficulties, when tackling this problem with bio-inspired algorithms, is handling larger time intervals (e.g., hours, or days). To address this issue, we propose a novel interestingness measure designed to limit the scope of time segments. In addition, we present experiments and results that illustrate the potential by application of the new interestingness measure.