Claudia Torres1, Samier Barguil2, Miguel Melgarejo2, Andrés Olarte3. 1. Laboratorio de Automática e Inteligencia Computacional, Facultad de Ingeniería, Universidad Distrital Francisco José de Caldas, Carrera 7 No. 40-53, Bogotá, Colombia. Electronic address: crtorresm@correo.udistrital.edu.co. 2. Laboratorio de Automática e Inteligencia Computacional, Facultad de Ingeniería, Universidad Distrital Francisco José de Caldas, Carrera 7 No. 40-53, Bogotá, Colombia. 3. Grupo de Modelamiento y Control de Sistemas Biológicos, Facultad de Ingeniería, Universidad Nacional de Colombia, Carrera 45 No. 26-85, Bogota, Colombia.
Abstract
OBJECTIVE: This article presents a model of a dengue and severe dengue epidemic in Colombia based on the cases reported between 1995 and 2011. METHODOLOGY: We present a methodological approach that combines multiresolution analysis and fuzzy systems to represent cases of dengue and severe dengue in Colombia. The performance of this proposal was compared with that obtained by applying traditional fuzzy modeling techniques on the same data set. This comparison was obtained by two performance measures that evaluate the similarity between the original data and the approximate signal: the mean square error and the variance accounted for. Finally, the predictive ability of the proposed technique was evaluated to forecast the number of dengue and severe dengue cases in a horizon of three years (2012-2015). These estimates were validated with a data set that was not included into the training stage of the model. RESULTS: The proposed technique allowed the creation of a model that adequately represented the dynamic of a dengue and severe dengue epidemic in Colombia. This technique achieves a significantly superior performance to that obtained with traditional fuzzy modeling techniques: the similarity between the original data and the approximate signal increases from 21.13% to 90.06% and from 18.90% to 76.83% in the case of dengue and severe dengue, respectively. Finally, the developed models generate plausible predictions that resemble validation data. The difference between the cumulative cases reported from January 2012 until July 2013 and those predicted by the model for the same period was 24.99% for dengue and only 4.22% for severe dengue. CONCLUSIONS: The fuzzy model identification technique based on multiresolution analysis produced a proper representation of dengue and severe dengue cases for Colombia despite the complexity and uncertainty that characterize this biological system. Additionally, the obtained models generate plausible predictions that can be used by surveillance authorities to support decision-making oriented to designing and developing control strategies.
OBJECTIVE: This article presents a model of a dengue and severe dengue epidemic in Colombia based on the cases reported between 1995 and 2011. METHODOLOGY: We present a methodological approach that combines multiresolution analysis and fuzzy systems to represent cases of dengue and severe dengue in Colombia. The performance of this proposal was compared with that obtained by applying traditional fuzzy modeling techniques on the same data set. This comparison was obtained by two performance measures that evaluate the similarity between the original data and the approximate signal: the mean square error and the variance accounted for. Finally, the predictive ability of the proposed technique was evaluated to forecast the number of dengue and severe dengue cases in a horizon of three years (2012-2015). These estimates were validated with a data set that was not included into the training stage of the model. RESULTS: The proposed technique allowed the creation of a model that adequately represented the dynamic of a dengue and severe dengue epidemic in Colombia. This technique achieves a significantly superior performance to that obtained with traditional fuzzy modeling techniques: the similarity between the original data and the approximate signal increases from 21.13% to 90.06% and from 18.90% to 76.83% in the case of dengue and severe dengue, respectively. Finally, the developed models generate plausible predictions that resemble validation data. The difference between the cumulative cases reported from January 2012 until July 2013 and those predicted by the model for the same period was 24.99% for dengue and only 4.22% for severe dengue. CONCLUSIONS: The fuzzy model identification technique based on multiresolution analysis produced a proper representation of dengue and severe dengue cases for Colombia despite the complexity and uncertainty that characterize this biological system. Additionally, the obtained models generate plausible predictions that can be used by surveillance authorities to support decision-making oriented to designing and developing control strategies.
Authors: Michael A Johansson; Nicholas G Reich; Aditi Hota; John S Brownstein; Mauricio Santillana Journal: Sci Rep Date: 2016-09-26 Impact factor: 4.379