Forecasting number of dengue patients using cellular automata model

This paper presents a novel forecasting model for dengue patient number using cellular automata (CA). The proposed model takes a number of people in each status of an epidemic model called SIER into consideration. In this respect, CA take a Genetic Algorithm (GA) to generate the factor weight chromosomes and Artificial Neural Network (ANN) to determine the probability of state transition `S´ to `E´ at time step t (P_t(s,e)). In addition, other related probabilities are obtained by expert knowledge; P(e,i) = 0.15 and P(i,s)=0.001. P(r,s) is determined by GA. These probabilities were used to calculate the cell number of each state at the next time step of CA. CA compute the fitness for one time step and repeat every time step finally to compute RMSE. For performance evaluation, 32 factors of dengue causes are used in the model. The dataset collected during 2005 to 2011 consisting of 359 weeks in which 287 and 72 are used to train and test the model, respectively. The results showed that the proposed model outperforms the compared ANN.