DYNAMIC ANALYSIS AND MONITORING OF AI BEHAVIOUR IN A MILITARY SIMULATION ENVIRONMENT USING MACHINE LEARNING TECHNIQUES IN THE ARMA 3 SIMULATOR

This dataset contains the complete behavioral telemetry results of 18 controlled simulation runs performed in Arma 3. The experiments were conducted under six experimental conditions resulting from the combination of three AI skill levels (0%, 50%, 100%) and two tactical configurations (VCOM OFF / VCOM ON). Each condition was replicated three times (R1–R3), resulting in a total of 18 independent runs. Each Excel file includes processed behavioral clustering outputs extracted from server-generated runtime logs. The dataset contains: Behavioral cluster durations (time_span) Sequential behavioral transitions Cluster identifiers Temporal segmentation data Enemy distance metrics (where applicable) The files represent direct outputs from the behavioral analysis pipeline and contain no manual annotation. All data were automatically extracted from Arma 3 server telemetry logs using a Python-based processing pipeline. This dataset enables full replication of the quantitative analysis presented in the associated article.