Enhancing intelligence source performance management through two-stage stochastic programming and machine learning...

Area of Science:

• Intelligence Studies
• Operations Research
• Machine Learning

Background:

• Human intelligence (HUMINT) source reliability is crucial for intelligence operations but is often unpredictable.
• Uncertainty in source behavior complicates resource allocation and tasking decisions.

Purpose of the Study:

• To develop a hybrid framework for managing HUMINT source performance under uncertainty.
• To optimize task allocation and mitigate risks associated with unpredictable source behavior.

Main Methods:

• Developed a hybrid framework combining Machine Learning (ML) and Two-Stage Stochastic Programming (TSSP).
• Utilized Extreme Gradient Boosting (XGBoost) and Support Vector Machines (SVM) for behavioral classification and reliability/deception prediction.
• Integrated predictive outputs as scenario probabilities into the TSSP model for optimized task allocation.

Main Results:

• Achieved 98% accuracy in behavioral classification; regression models yielded R-squared scores of 93% (reliability) and 81% (deception).
• The hybrid framework reduced expected tasking costs by 16.8% and improved mission success rates by 19.3% compared to deterministic methods.

Conclusions:

• Scenario-based probabilistic planning significantly outperforms static heuristics in managing HUMINT operational uncertainty.
• The developed framework shows promise for enhancing HUMINT operations, with further validation needed through field data.