Over recent decades, the world has experienced a growing demand for and reliance upon unmanned aerial systems (UAS) to perform a broad spectrum of applications to include military operations such as surveillance/reconnaissance and strike/attack. As UAS technology matures and capabilities expand, especially with respect to increased autonomy, acquisition professionals and operational decision makers must determine how best to incorporate advanced capabilities into existing and emerging mission areas. This research seeks to predict which autonomous UAS capabilities are most likely to emerge over the next 20 years as well as the key challenges for implementation for each capability. Employing the Delphi method and relying on subject matter experts from operations, acquisitions and academia, future autonomous UAS mission areas and the corresponding level of autonomy are forecasted. The study finds consensus for a broad range of increased UAS capabilities with ever increasing levels of autonomy, but found the most promising areas for research and development to include intelligence, surveillance, and reconnaissance (ISR) mission areas and sense and avoid and data link technologies.

Marijuana is the most commonly abused drug for military personnel tested at the Air Force Drug Testing Laboratory. A publicly available dataset of drug use, personality trait scores and demographic data was modeled with logistic regression, decision tree and neural network models to determine the extent to which marijuana use can be predicted using personality traits. While the logistic regression model had lower performance than the neural network model, it matched the sensitivity of prior work (0.80), achieved a high level of significance (p < 0.05) and yielded valuable inferences. It implied that younger, less educated individuals who exhibit sensation-seeking behavior and are open to experience tend to be at higher risk for THC use. A method for performing an iterative multidimensional neural network hyperparameter search is presented, and two iterations of a 6-dimensional search were performed. Metrics were used to select a family of 8 promising models from a cohort of 4600 models, and the best NN model’s 0.87 sensitivity improved upon the literature. The model met an f1 overfitting threshold on the test and holdout datasets, and an accuracy sensitivity analysis on a holdout-equivalent dataset yielded a 95% CI of 0.86 ± 0.04. These results have the potential to increase the efficacy of drug prevention and intervention programs.

Current food crisis predictions are developed by the Famine Early Warning System Network, but they fail to classify the majority of food crisis outbreaks with model metrics of recall (0.23), precision (0.42), and f1 (0.30). In this work, using a World Bank dataset, classical and neural network (NN) machine learning algorithms were developed to predict food crises in 21 countries. The best classical logistic regression algorithm achieved a high level of significance (p < 0.001) and precision (0.75) but was deficient in recall (0.20) and f1 (0.32). Of particular interest, the classical algorithm indicated that the vegetation index and the food price index were both positively correlated with food crises. A novel method for performing an iterative multidimensional hyperparameter search is presented, which resulted in significantly improved performance when applied to this dataset. Four iterations were conducted, which resulted in excellent 0.96 for metrics of precision, recall, and f1. Due to this strong performance, the food crisis year was removed from the dataset to prevent immediate extrapolation when used on future data, and the modeling process was repeated. The best “no year” model metrics remained strong, achieving ≥0.92 for recall, precision, and f1 while meeting a 10% f1 overfitting threshold on the test (0.84) and holdout (0.83) datasets. The year-agnostic neural network model represents a novel approach to classify food crises and outperforms current food crisis prediction efforts.

Recent advances in large language models and retrieval-augmented generation, a method that enhances language models by integrating retrieved external documents, have created opportunities to deploy AI in secure, offline environments. This study explores the feasibility of using locally hosted, open-weight large language models with integrated retrieval-augmented generation capabilities on CPU-only hardware for tasks such as question answering and summarization. The evaluation reflects typical constraints in environments like government offices, where internet access and GPU acceleration may be restricted. Four models were tested using LocalGPT, a privacy-focused retrieval-augmented generation framework, on two consumer-grade systems: a laptop and a workstation. A technical project management textbook served as the source material. Performance was assessed using BERTScore and METEOR metrics, along with latency and response timing. All models demonstrated strong performance in direct question answering, providing accurate responses despite limited computational resources. However, summarization tasks showed greater variability, with models sometimes producing vague or incomplete outputs. The analysis also showed that quantization and hardware differences affected response time more than output quality; this is a tradeoff that should be considered in potential use cases. This study does not aim to rank models but instead highlights practical considerations in deploying large language models locally. The findings suggest that secure, CPU-only deployments are viable for structured tasks like factual retrieval, although limitations remain for more generative applications such as summarization. This feasibility-focused evaluation provides guidance for organizations seeking to use local large language models under privacy and resource constraints and lays the groundwork for future research in secure, offline AI systems.

Physical fitness is a pillar of U.S. Air Force (USAF) readiness and ensures that Airmen can fulfill their assigned mission and be fit to deploy in any environment. The USAF assesses the fitness of service members on a periodic basis, and discharge can result from failed assessments. In this study, a 21-feature dataset was analyzed related to 223 active-duty Airmen who participated in a comprehensive mental and social health survey, body composition assessment, and physical performance battery. Graphical analysis revealed pass/fail trends related to body composition and obesity. Logistic regression and limited-capacity neural network algorithms were then applied to predict fitness test performance using these biomechanical and psychological variables. The logistic regression model achieved a high level of significance (p < 0.01) with an accuracy of 0.84 and AUC of 0.89 on the holdout dataset. This model yielded important inferences that Airmen with poor sleep quality, recent history of an injury, higher BMI, and low fitness satisfaction tend to be at greater risk for fitness test failure. The neural network model demonstrated the best performance with 0.93 accuracy and 0.97 AUC on the holdout dataset. This study is the first application of psychological features and neural networks to predict fitness test performance and obtained higher predictive accuracy than prior work. Accurate prediction of Airmen at risk of failing the USAF fitness test can enable early intervention and prevent workplace injury, absenteeism, inability to deploy, and attrition.

Given the emergence of China as a political and economic power in the 21st century, there is increased interest in analyzing Chinese news articles to better understand developing trends in China. Because of the volume of the material, automating the categorization of Chinese-language news articles by headline text or titles can be an effective way to sort the articles into categories for efficient review. A 383,000-headline dataset labeled with 15 categories from the Toutiao website was evaluated via natural language processing to predict topic categories. The influence of six data preparation variations on the predictive accuracy of four algorithms was studied. The simplest model (Naïve Bayes) achieved 85.1% accuracy on a holdout dataset, while the most complex model (Neural Network using BERT) demonstrated 89.3% accuracy. The most useful data preparation steps were identified, and another goal examined the underlying complexity and computational costs of automating the categorization process. It was discovered the BERT model required 170x more time to train, was slower to predict by a factor of 18,600, and required 27x more disk space to save, indicating it may be the best choice for low-volume applications when the highest accuracy is needed. However, for larger-scale operations where a slight performance degradation is tolerated, the Naïve Bayes algorithm could be the best choice. Nearly one in four records in the Toutiao dataset are duplicates, and this is the first published analysis with duplicates removed.