J. Scott Semel
Senior Principal Analyst
Dr. Semel has 14 years experience in projects using operations research, bayesian analysis, and machine learning. He is currently analyzing cyber threats to DoD networks. He was a co-developer of modeling of pre-activation survivability during CNA presented to 74th MORSS. He supported exercise Terminal Fury with Extend Modeling of IOJMEMS factors. He supported creation of new risk assessment tool for CNSS with prioritization of NIST controls against multiple adversaries for CND. He created threat rankings for agent resource allocation using Logical Decisions software. He performed validation testing of multiple network models using Infer, ORA, Netica, and Renoir. He performed risk and threat assessments through synthesis of information across intelligence disciplines for NC2 networks. He performed decision analysis on new hypothesis generation techniques for applications in data mining. He built a data mining tool in MATLAB to classify records with mixture of text and numeric data and training on relevant records and testing multiple classifiers on each binarization technique. He developed stand-alone geospatial data mining software using MATLAB with C-compiler, fuzzy logic, statistics, optimization, and mapping toolboxes and developed classification techniques based on association rules, artificial neural networks, decision trees, and logistic regression. He directed summer research of AFIT and NPS graduate students for new methods in effects-based targeting. He investigated the use of hidden Markov models versus neurodynamic programming to code KEDS world event data and modeled lines of communications vulnerabilities for various military exercises. He directed development of a time-expanded quickest transshipment algorithm to model network interdiction of military convoys. He developed method and code using deconvolution of brain potential maps with statistical and decision analysis on patients with mild traumatic brain injury versus normal patients for his dissertation and adapted the method to solve for source location of aerial contaminant release through the backward heat equation. He tested different wavelet compression algorithms for EEG data transmission across the Internet, and tested the value-added in ROI of additional electrode sampling.
Ph.D., Applied Mathematics, University of Louisiana at Lafayette, 1997
M.S., Applied Mathematics, University of Louisiana at Lafayette, 1992
B.A. Mathematics, University of Southern California, 1988
Selected Publications and Presentations
G. Carstens, M. King, S. Semel, M. Horejs, “Methodology for Enhancing the Pre-Activation Survivability (PAS) Calculation in the Effects Expectancy (EE) Framework, “presented at 74th MORSS, 2006.
R.D. Sidman, J.S. Semel, T.D. Lagerlund, and M.R. Ford, “The Effect of Reference-Electrode Choice on the Spatial Resolution of Topographical Potential Maps in the Discrimination of Deep Cerebral Sources,” J. Neuroscience Methods, 68 (1996) 175-184.
R.D. Sidman, C.H. Chu, T.D. Lagerlund, and J.S. Semel, “Computing Environment and Tools for Network-Based Interactive Neuroscience Data Visualization,” (abstract) J. Clin. Neurophysiology, 13 (1996) 440-441.
J.S. Semel, R.D. Sidman, L. Ke, and T.D. Lagerlund, “The Effect of Noise in Scalp-Recorded EEG and EP Data in Approximating Voltages on the Cortical Surface,” (abstract) J. Clin. Neurophysiology, 13, (1996) 453.