Avila University /ˈævɪlə/ is a private Roman Catholic university in Kansas City, Missouri. It is sponsored by the Sisters of St. Joseph of Carondelet and offers bachelor's degrees and master's degrees. Its 13 buildings are situated on a campus of 50 acres (20.2 ha) in Kansas City. The school enrolled 1,527 students in 2019.
Theoretical frameworks suggest that romantic relationship satisfaction may influence sexual communication self-efficacy, which involves the confidence to discuss personal sexual health goals and plays a key role in HIV prevention among partnered adolescent males who have sex with males (MSM). The current study examined romantic relationship satisfaction as a correlate of sexual communication self-efficacy in a sample of partnered adolescent MSM. Participants (n = 50) completed self-report measures including romantic relationship satisfaction and sexual communication self-efficacy. Domains of sexual communication self-efficacy included: sexual history, contraception, condom negotiation, negative sexual messages, and positive sexual messages. Multivariable linear regression models were conducted to examine study aims. Relationship satisfaction was positively associated with self-efficacy in communicating about sexual history, contraception, and negative and positive sexual messages, with effect sizes ranging from moderate to large. These findings shed light on pathways often assumed in HIV intervention programs, emphasizing how romantic relationships are associated with adolescents' confidence in discussing sexual health with partners. Interventions aiming to improve sexual communication may be more effective if they include content on enhancing romantic relationship satisfaction.
Existing supply chain management systems typically involve siloed, centralised applications with no true visibility, predictive intelligence, and flexibility to withstand any disruption in the chain. This study introduces a Blockchain-Enabled Digital Twin Framework for Intelligent Demand Forecasting and Supply Chain Resilience (BDT-DFR) to tackle these challenges. The framework to materialize the secure and transparent supply chain ecosystem is a combination of blockchain technology, digital twin modelling, demand forecasting using machine learning, and assessment functions. The blockchain technology is decentralized and inaccessible, which guarantees trust, transparency, correctness, and traceability for suppliers, manufacturers, logistics service providers, distributors, and retailers. Digital twins automatically update the physical entities in the supply chain with their virtual counterparts, allowing real-time tracking and modeling of all operations. The module for intelligent forecasting involves the use of Long Short-Term Memory (LSTM) models, XGBoost models, and Random Forest models to forecast future fluctuations and inventory needs, transit delays, and possible supply disruptions. Additionally, by utilizing a resilience management layer, the supply chain robustness will be profiled based on disruption recovery capability, supplier reliability, logistics flexibility, and operational continuity indicators. Automated smart contracts ensure accurate compliance monitoring, replenish inventory, and confirm transactions, automating workflows and boosting efficiency. A series of experiments with the benchmark supply chain datasets proves that the proposed BDT-DFR offers an accuracy of 98.8%, a precision of 98.4%, a recall of 98.1%, and an F1 score of 98.2%, as opposed to the conventional blockchain-based and digital twin-based approach of supply chain management.
This paper examined at how an advertising budget can be optimized using a machine learning architecture in the case of traditional advertising channels, i.e., television, radio, and newspaper advertising. The key task was to model and test a predictive model that could quantify the contribution of each channel to sales and thus help with more evidence-based budget allocations. It adopted a computational cross-sectional design that rested on secondary data analysis. In the study, a publicly available advertising dataset of 200 observations with four variables was used in the form of television advertising budget, radio advertising budget, newspaper advertising budget, and sales. Descriptive statistics, Pearson correlation, and multiple linear regression were employed to analyze the data. The results had indicated that the general regression model was statistically significant, F(3, 196) = 570.271, p < .001, and explained 89.7% of the variance in sales (R² = .897; Adjusted R² = .896). Television advertising emerged as the strongest predictor of sales (β = .753, p < .001), followed by radio advertising (β = .536, p < .001), while newspaper advertising did not make a statistically significant independent contribution (β = − .004, p = .860) Such findings suggest that the traditional advertising mediums vary significantly in terms of their marginal performance and the analysis of budget allocation can be done based on machine learning as opposed to intuition-driven planning to offer a more dependable foundation of budget allocation. The research adds to the literature in that machine learning-based allocation studies have been limited to digital advertising setting so far and cannot be extended to a traditional media setting. In practice, it provides managers with a clear system of priorities on the way to concentrate on high-impact channels and enhance accountability of marketing spending activities. The researchers state that information-grounded allocation has the potential to enhance the advertisement decision-making process and infers basis of future research based on richer, longitudinal, and multi-channel data sets.
Every year, industrial accidents kill hundreds of thousands of workers and injure millions more. Machine learning offers a path toward predicting incident severity before things go wrong. Most published approaches, however, share two serious weaknesses. They measure success with accuracy on imbalanced data, which hides failures on the very incidents that matter most, and they produce predictions that nobody can explain or act on. We tackle both problems. Using the IHM Stefanini industrial safety database (425 real-world incident records spanning mining, metallurgy, and manufacturing), we benchmark six classifiers (Logistic Regression, Random Forest, XGBoost, LightGBM, SVM, and MLP) under a rigorous 10-fold stratified cross-validation protocol. Rather than picking one imbalance strategy, we test two head-to-head. SMOTE-based oversampling and cost-sensitive class weighting each have different strengths, and the results are model-specific. SMOTE helps Random Forest most. Gradient boosting methods benefit more from cost-sensitive weighting. Left untreated, Random Forest misses 96% of high-severity incidents despite looking reasonable on standard metrics. On explainability, we apply SHAP to open the black box. Three features consistently drive severity predictions: the independently assessed potential accident level, the industry sector (mining incidents carry disproportionately high risk), and the mechanical risk macro-category. These findings align with established occupational safety theory and point directly toward preventive action. A cross-domain evaluation against the OSHA construction injury database tests whether learned risk patterns hold beyond their origin sector. This work bridges the gap between raw predictive performance and the kind of interpretable, decision-ready safety intelligence that EHS practitioners can actually use.