Autoregressive integrated moving average with semantic information: An efficient technique for intelligent prediction of dengue cases

Herein, we proposed a novel approach, AutoRegressive Integrated Moving Average with Semantic Information (ARIMAS), an extension of the ARIMA model that incorporates a knowledge base about dengue outbreaks. This extension considerably improves the analysis capabilities of traditional time-series forecasting models. ARIMAS differentiates itself by integrating semantic information from external knowledge bases, enabling a comprehensive understanding of the relationships between data points using an ontological model. This approach extracts related data from a dataset to enhance the autoregressive technique beyond ARIMA's reliance on historical time-series data alone. We evaluate ARIMAS using a provincial dengue dataset from Thailand, comparing its performance with those of state-of-the-art methods: ARIMA, simple exponential smoothing (SES), and Holt's exponential smoothing (HES). The results demonstrate that ARIMAS achieves an average mean square error of 4.88E+05, outperforming ARIMA (5.04E+05), SES (7.60E+05), and HES (7.67E+05). ARIMAS represents a promising advancement in time-series forecasting for dengue outbreaks by addressing the key limitations of ARIMA models. By combining statistical analysis with domain-specific knowledge, ARIMAS can provide more accurate, interpretable, and contextually relevant predictions, ultimately supporting effective public health interventions. This advancement in prediction accuracy highlights the prediction effectiveness of ARIMAS, particularly in unusual situations that may exhibit short, nonlinear trends. This enhanced method can eliminate nonlinearity through outlier reduction and is thus a robust, superior tool for time-series analysis that overcomes the constraints of traditional models.