Page 671 - NGTU_paper_withoutVideo

Modern Geomatics Technologies and Applications

2.3.5. Calibration and validation of the proposed agent-based model: Due to the indeterminacy as well as lack of some
parameters’ values (e.g., the number of mosquitoes, the infected percentage of mosquitoes, and the gravid percentage of
mosquitoes) at the beginning of the model in the study area, the simulated model was executed for different initial values of
these parameters. The results of the model (the monthly average number of infected human agents in 100 times) were compared
with the number of actual malaria cases in each month of 2013 according to the Root Mean Square Error (RMSE). The values
considered for the aforementioned unknown parameters were determined when the lowest value was obtained for the RMSE. In
that way, the simulated agent-based model was calibrated and validated.

3. Results and discussion
The main output of the proposed agent-based model is the number of the infected human agent in each month; based on
this output, the calibration and validation of the model were performed. In Fig.3, the proposed agent-based model’s interface
was illustrated.

Fig. 3. The interface of the proposed agent-based model in NetLogo 6.0.4 platform.

According to the comparison of the model’s results (i.e., the average number of infected human agents in each month)
with the actual observed value (i.e., the monthly number of malaria cases) (Fig.4) based on the RMSE, the proposed agent-based
model was calibrated as well as validated. The value of 6.853 was the best RMSE value obtained (i.e., the lowest value amongst
all RMSE values). Besides, as the result of the model calibration, the initial number of mosquito agents was considered 1650, of
which about 9% were considered infected and approximately 2% were considered as gravid mosquito agents.
60
Number of infected people 40 Predicted value
50

30
20
Actual observed value
10
0

Month
Fig. 4. Comparison of the actual observed and the predicted number of infected people.

5