Predicting dengue outbreaks using approximate entropy algorithm and pattern recognition

The prediction of dengue outbreaks is a critical concern in many countries. However, the setup of an ideal prediction system requires establishing numerous monitoring stations and performing data analysis, which are costly, time-consuming, and may not achieve the desired results. In this study, we developed a novel method for predicting impending dengue fever outbreaks several weeks prior to their occurrence.

By reversing moving approximate entropy algorithm and pattern recognition on time series compiled from the weekly case registry of the Center for Disease Control, Taiwan, 1998-2010, we compared the efficiencies of two patterns for predicting the outbreaks of dengue fever.

The sensitivity of this method is 0.68, and the specificity is 0.54 using Pattern A to make predictions. Pattern B had a sensitivity of 0.90 and a specificity of 0.46. Patterns A and B make predictions 3.1 ± 2.2 weeks and 2.9 ± 2.4 weeks before outbreaks, respectively.

Combined with pattern recognition, reversed moving approximate entropy algorithm on the time series built from weekly case registry is a promising tool for predicting the outbreaks of dengue fever.