Ph.D. Thesis Defense

By

Mahmoud A. Abdelaal

(Advisor: Prof. Dimitri N. Mavris)

9:00AM, Wednesday, July 06, 2016

Weber Space Science and Technology Building (SST-II)

Collaborative Visualization Environment (CoVE)

CLUSTER ANALYSIS FOR DECISION-POINT RECOGNITION

ABSTRACT:

Utilizing the latest in technology, today’s military is engaged in complex conflicts and operations across the globe.  These disparate operations can occur simultaneously and within the same theater.  Joint operational planning processes are often characterized as data, labor and time intensive, and courses of action (COA) must be planned and executed within the enemy’s decision making cycle.  This represents a computational burden that scales dramatically with increasing numbers of systems and actors within the modern battlespace.  The most time consuming part of the planning process is COA analysis and wargaming.  This highlights the research objective developing a methodology to aid military planners by utilizing a new process for analyzing and evaluating COA alternatives.

By analyzing the effect of the interaction of the terrain, systems, and actors within the battlespace, it is hypothesized that leading indicators and metrics associated with a specific battlefield configuration will allow the identification of “rules of thumb” to follow.  Wargaming different scenarios will result in a heterogeneous database to analyze because the variations in plan that will be simulated will have an effect on the protraction or brevity of the simulated operation.  Furthermore, this heterogeneous data is temporal in nature, where the metrics are time series associated with the occurrences on the simulated battlefield.  Using these data constraints to develop the methodology, Hierarchical Agglomerative Clustering, using the Dynamic Time Warping algorithm for similarity measurement were utilized.  Additionally, a novel method for cluster validation was created to establish relative value between different linkage algorithms using the similarity height of dendrograms and the statistical significance of the within cluster outcomes. 

The primary contribution of this thesis is offering of enhanced COA analyses of wargame data allowing the identification of decision points and heuristics.  As a result of the formulation and the experiments, this work has aided in the creation of a cluster validation index that considers inner cluster similarity and the statistical significance of the outcome of scenarios.  Additionally, this work has contributed by the determination of the effect of scaling of complexity in the number of elements and in the complexity of the underlying wargame on the selection and performance of linkage algorithms, HSig Index, identification of decision points and heuristics.

Committee Members:

Professor Dimitri N. Mavris

Professor Daniel P. Schrage

Professor Graeme J. Kennedy

Dr. Jean Charles Domercant

Dr. K. Daniel Cooksey