Time-series models for border inspection data.

Area of Science:

• Biosecurity and Invasive Species Management
• Statistical Modeling and Time Series Analysis
• Agricultural and Trade Inspection Data

Background:

• Current methods classify shipments based on historical data, but lack real-time dynamic risk assessment.
• Effective cargo inspection requires models that adapt to changing risk levels and focus efforts efficiently.
• Distinguishing systematic from random pest detection patterns is crucial for regulatory response.

Purpose of the Study:

• To propose a novel dynamic modeling approach for inspection data, interpreting invasive pest detection patterns.
• To develop a time-series-based methodology for real-time detection of high-risk periods in cargo.
• To compare the predictive performance of models with varying degrees of serial dependence.

Main Methods:

• Developed a dynamic time-series approach for analyzing cargo inspection data.
• Compared three statistical models: independence model, Markov chain, and hidden Markov model.
• Evaluated model predictive performance using Receiver Operating Characteristic (ROC) and leakage curves.

Main Results:

• The dynamic time-series models demonstrated improved interpretation of invasive pest detection patterns.
• Models accounting for serial dependence (Markov chains) showed potential for enhanced risk assessment.
• Comparative analysis using ROC and leakage curves provided insights into model effectiveness.

Conclusions:

• The proposed dynamic modeling approach offers a more useful interpretation of inspection data for invasive pests.
• Time-series analysis and models incorporating serial dependence are valuable for real-time biosecurity risk management.
• This methodology can aid regulatory organizations in more effectively targeting inspection efforts.