Suratthani Rajabhat University

This study compares machine learning and econometric approaches for forecasting agricultural export values in volatile global markets, examining predictive accuracy and economic interpretability trade-offs. Monthly data from January 2014 to December 2023 were analyzed using five models: Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM), Hybrid ANN-LSTM, Ordinary Least Squares (OLS), and Autoregressive Distributed Lag (ARDL). Key predictors included durian, mangosteen, and longan export values/volumes, plus China's GDP. Performance evaluation used MAE, RMSE, MAPE, and R² metrics with systematic hyperparameter optimization through grid search and 5-fold cross-validation. ANN achieved the highest absolute accuracy (MAE: 1,684,667,401.55; RMSE: 2,602,671,952.28), while Hybrid ANN-LSTM delivered superior relative accuracy (MAPE: 1.58%). ARDL demonstrated exceptional explanatory power (R²=0.83) for structural economic relationships. China's GDP emerged as the strongest determinant across all models. Longan export value showed contrasting effects between approaches, positive in machine learning models versus negative in econometric models, reflecting different paradigmatic interpretations of market substitution dynamics. This research introduces the first comprehensive comparative framework integrating advanced hybrid neural networks with traditional econometric methods for multi-commodity agricultural forecasting, addressing cross-commodity substitution effects previously unexplored while offering complementary perspectives for both predictive accuracy and economic policy interpretation.

In Thailand, the incidence of falls is higher in rural areas compared to urban regions, primarily due to environmental factors. While studies have examined fall risk factors, few have investigated home environments in rural coastal areas and their impact on falls among older adults. This study sought to investigate the relationship between home environmental risk factors and falls among older adults living in coastal rural communities in Thailand, contributing to targeted fall prevention strategies. This cross-sectional study enrolled a total of 60 older adults, aged between 60 and 93 years, residing in rural coastal communities of Surat Thani Province. Participants were required to be Thai nationals living in their own homes, able to communicate independently with researchers, and willing to provide informed consent. Demographic and home environmental data were collected through structured interviews and corroborated by on-site observations of each participant’s home environment. A fall was operationally defined as having experienced at least one falling incident during the past six months. Descriptive statistics were applied to summarize participants’ characteristics, while Chi-square and Fisher’s exact tests were utilized to examine the association between socioeconomic and environmental factors and fall variables. Within rural coastal communities, 17% (95% CI: 7.2%-26.1%) of older adults reported experiencing at least one fall during the preceding six months. Bivariate analyses were conducted to explore factors associated with fall occurrence. In these analyses, being unemployed was significantly associated with a higher prevalence of falls (p = 0.002), whereas having a mosquito net installed in the bedroom was associated with a reduced risk of falling (p = 0.031). Falls were relatively common among older adults living in rural coastal areas, and two such socioeconomic and environmental factors demonstrated significant associations with fall occurrence. These findings suggest that initiatives aimed at addressing unemployment-related vulnerabilities and promoting practical home environment modifications may be beneficial for reducing fall risk in these communities, although further research is required to confirm causal pathways and evaluate the effectiveness of such interventions.

ABSTRACT This study explored the factors influencing residents’ intentions to support Muslim-friendly tourism, focusing on attitudes, avoidance behaviors, image perceptions, knowledge, and experiences. The findings revealed that residents’ attitudes, avoidance responses, and perceived image of Muslim-friendly tourism significantly impacted their intentions, while knowledge and experiences did not have a significant predictive effect. By applying Cumulative Prospect Theory and utilizing fsQCA analysis, the study identified four key causal configurations that shape residents’ perceptions of Muslim-friendly tourism. It demonstrated that positive attitudes and avoidance responses are crucial factors. Notably, loss factors, such as Islamophobia, exerted a greater influence than gain factors, underscoring the need to address negative perceptions to enhance local acceptance of Muslim-friendly tourism. Additionally, a cross-cultural comparison indicated that US residents generally possess more positive perceptions, greater familiarity, and stronger support for Muslim-friendly tourism compared to South Korean residents, who tend to be more cautious and exhibit higher levels of avoidance.

Alzheimer's disease (AD) is a persistent neurodegenerative disorder known for its impact on memory and cognitive abilities. An early and accurate diagnosis, particularly through neuroimaging (especially MRI), is crucial for implementing effective therapeutic strategies and management. This article evaluates various deep learning-based architectures for AD automation using an open-source Kaggle dataset. We develop and compare a Simple CNN, a Deep CNN, MobileNetV2, and EfficientNetB0 for accuracy of classification, convergence on the loss function value, and generalisation ability. The confusion matrix, precision, recall, and F1-score, together with other evaluation metrics for different categories (Non-Demented; Very Mild Demented; Mild Demented and Moderate Demented), are considered. We observe the bias-variance trade-off in model complexity and performance, with Deep CNN showing the highest accuracy while being computationally efficient. This study illustrates that deep learning is beneficial in clinical diagnosis and offers empirical support for further exploration of efficient neural architectures for identifying neurodegenerative diseases.

The combination of pasteurization and Sa-tay marination represents a practical and applicable strategy for enhancing the microbial safety, sensory quality, and shelf life of oysters, thereby supporting both public health and the seafood industry. This study aimed to investigate the effects of pasteurization and Sa-tay marination on quality changes in oyster meat after inoculation with test pathogenic bacteria. Results showed that pasteurization at 80oC for 8 minutes completely eliminated an initial inoculum of pathogenic bacteria (108 CFU/g), including Escherichia coli, Salmonella Typhimurium, Vibrio cholerae, V. parahaemolyticus, and V. vulnificus. The combination of pasteurization and Sa-tay marination effectively eliminated E. coli, S. Typhimurium, V. cholerae, V. parahaemolyticus, and V. vulnificus in oyster meat. However, the total volatile base (TVB) value slightly increased after 9 and 12 days of storage. In contrast, trimethylamine nitrogen (TMA-N) was not detected. In summary, pasteurization is a highly effective method for reducing pathogenic bacteria in oyster meat. Without inoculation with pathogenic bacteria, oyster meat marinated with Sa-tay showed improved safety, resulting in lower microbial counts. When combined with Sa-tay marination, this approach also supports extended shelf life. This technique offers a practical solution for seafood processors aiming to improve product safety and commercial viability.

The prediction of student performance is a key area for researchers aimed at enhancing student competence and improving educational outcomes. This study focuses on identifying the most accurate machine learning techniques for predicting the performance of preservice teachers (PSTs) by analyzing their course GPA and cumulative grades. The dataset was collected from universities in Cambodia between 2021 and 2024. We used ANOVA for feature selection and evaluation of various machine learning algorithms by applying accuracy, precision, recall, and F1-score. We found that SVM and KNN achieved the highest accuracy, equal to 0.98 with a confidence interval of $[0.94-1.00]$, in predicting PSTs’ performance. The findings of this study could help educators develop effective strategies for anticipating and improving the performance of PSTs by providing timely support and feedback.