Modeling Dengue Cases and Online Search Behavior for Prediction Models

Dengue continues as a pressing concern in the public health due to its widespread prevalence. Disease surveillance is challenging public health agencies to determine the number and distribution of cases as well as severity of disease in the community. Technologies like social media have found utility to gather internet search data that provides support to the public for their information needs. Internet search data were found to be capable of tracking dengue-related activities to support surveillance. Various statistical methods have been used to predict the disease outbreaks including dengue. Regression analysis, based on time series data, revealed that dengue cases in the select cities are increasing over the years. Metrics such as mean average error, mean squared error and root mean-square deviation were calculated to test the accuracy of the predictive model. Exponential smoothing reveals to be the best model for forecasting, resulting in low mean values of the accuracy metrics. Assessing the model accuracy to predict dengue cases and “dengue” online search behavior may aid relevant stakeholders improve the design of early warning systems on dengue surveillance. Further research, extends to explore other sources of internet search data, i.e., social media which could potentially model disease spread from geographic locations.