Alert-Driven Pattern Mining in Large-Scale Road Traffic Management

The immense flood of alerts that is constantly produced in large-scale control systems (LSCS) and particularly in road traffic management (RTM), represents a substantial challenge for efficient and safe operation. Although research for reducing alert floods exists since decades, mining of appropriate alert patterns as the ultimate means to cope with alert quantity is especially challenging since relationships between alerts are commonly unknown, due to heterogeneity, size, and evolutionary nature. In search of the holy grail for dealing with alert floods, i.e., exploiting relationships between alerts, this paper contributes an alert-driven pattern mining approach, based on a hybrid, multi-objective evolutionary algorithm. This approach is unique in that first, pattern coverage is maximized, ensuring that each alert occurrence is pinned down within a pattern, thus allowing to reason about all underlying relationships for the whole alert log data. Second, pattern frequency is leveraged, ensuring that both frequent as well as rare patterns are found, thus allowing for alert flood reduction in regular as well as exceptional and possibly critical cases. Based on real-world log data in the area of RTM, the applicability of our approach is demonstrated, complemented by a comparative evaluation.