Neural parameter calibration for dengue outbreak forecasting

Comprehensive Summary

This research presented by Viet Pham et al. examined the use of mathematical models on anticipating Dengue Fever (Aedes mosquito-borne) outbreaks in Columbia, Peru, Puerto Rico, Vietnam, the Philippines, and Cambodia over a period of 48 weeks. Ordinary differential equations (ODEs, n = 7) were utilized to track patient health statuses as well as parameter estimation techniques such a traditional Markov chain Monte Carlo (MCMC) and machine learning Neural parameter calibration (NPC). The results showed that the computation time of NPC was significantly faster than MCMC with an average of 400 seconds compared to 2600-3000+ seconds respectively, approximately 7-8 times faster. In terms of accuracy, NPC was similar to and produced results even more accurate than MCMC. Of the most important factors contributing to the outbreak, those identified were rates of mosquito death, rates of human recovery, and transmission rates between humans and mosquitos. Viet Pham et al. acknowledged the need for weather data in future studies and the possible application of this technology on other diseases and their respective outbreaks.

Outcomes and Implications

Artificial intelligence estimates on disease outbreaks could become a critical prevention and treatment tool for governments battling epidemics with increased needs for rapid public health responses. Neural Parameter Calibration (NPC) has suggested that artificial intelligence has equal, if not, better accuracy in estimating outbreaks while being significantly faster than traditional statistics applications. Epidemiological modeling has the potential to be advanced by machine learning and take part in the effort to heal and protect communities worldwide from these contagious and life-threatening diseases.