Title: Improving the Risk-Scoring Algorithm of Marine Cargo Containers (Applications of Approximate Boolean Classifiers to Marine Container Security)
Speaker: Richard Hoshino, Canada Border Services Agency
Date: Monday, November 13, 2006 2:00pm
Location: CoRE Bldg, CoRE A, Room 301, Rutgers University, Busch Campus, Piscataway, NJ
Sponsors: The DIMACS Port of Entry Inspection Algorithms Project and the DIMACS/CAIT Laboratory for Port Security
Of the 1.8 million marine cargo containers arriving in Canada each year, only a small percentage are physically inspected for contraband and other undesirable goods. The ultimate question is, "which containers do we inspect"? Similar to the Automated Targeting System at the U.S. Customs and Border Protection, our organization uses an automated commercial risk-assessment system to identify the highest-risk containers. In this talk, we discuss how Logistic Regression and Approximate Boolean Classifiers can be used to improve CBSA's risk-scoring algorithm, by minimizing the percentages of false positives (clean containers incorrectly targeted for examination) and false negatives (dirty containers incorrectly released without examination). The effectiveness of our proposed risk-scoring algorithm is measured by calculating the area under the Receiver Operating Characteristic (ROC) Curve.