Dicluster: Differential Arrangements of Event Intervals Biclustering for Task Collaboration Safety Analysis

Abstract

The solution and application of single-pilot operations (SPOs) for commercial aircraft are an active area of research. While improving operational efficiency, safety issues occur with the introduction of the ground operator and enhanced air-ground interaction, especially in the situation of SPO air-ground multitask collaboration. Traditional safety methods are ineffective in discovering hazard patterns implicated in SPO task-function time interval data. Aiming at the above problem, an efficient time-interval-related pattern (TIRP) mining algorithm, Dicluster, is proposed to mine maximal differential biclusters of event interval arrangements from an SPO task-function matrix. The proposed algorithm addresses the deficiencies of existing competitors by employing several techniques, such as directed acyclic weighted graph for candidate generation, and pruning and output strategies for efficient time and memory consumption. Experiments on different scales of real-world datasets indicate that Dicluster is more efficient than other methods, and an experiment under typical SPO surface scenarios shows that Dicluster is practical to assist in the safety analysis of SPO task collaboration.