Page 46 - ASBIRES-2017_Preceedings

A TIME SERIES MODEL TO FORECAST DENGUE FEVER INCIDENCES IN KURUNEGALA DISTRICT

Using the above summary SARIMA 5 CONCLUSION
(0,1,0) (1,1,1)6 was identified as the suitable The main limitation was Dengue patient
model. count were available for only recent seven

Table 2: Final parameter estimation of years. The descriptive data such as gender,
SARIMA (0,1,0) (1,1,1)6 age and other categories were limited to
2014,2015,2016, since the data were not
SMA1 available for the rest of the period. Due to
Coefficients -1 the inconvenience of obtaining climatic data
Standard error 0.2191 such as rain fall, humidity and temperature
climatic factors were not used to create a

The auto arima function of R software model. The Dengue patient count in
suggested that the ARIMA (1,0,2) is the Puttalam district was smaller compare to the
suitable model for this data. Also, the Kurunegala district. Therefore, there were
conditions were checked.
no pattern to identify the parameters of the
Table 3: Accuracy measures model in Puttalam district. Hence, the model
was only fitted for Kurunegala district.
SARIMA ARIMA
(0,1,0) (1,1,1)6 (1,0,2) According to the final results two
models were suggested. They are SARIMA
MAPE 39.8% 44.06% (0,1,0) (0,1,1)6 and ARIMA (1,0,2). It was
AIC 857.45 945.57 identified that SARIMA (0,1,0) (0,1,1)6 is
BIC 862.03 957.48
the suitable model for the data with lowest
AIC and MAPE value. The number of
Referring to the above table SARIMA
(0,1,0) (0,1,1)6 contains lowest Akaike Dengue patients were forecasted using the
Information Criterion(AIC), Bayesian final model for the next six months.
Information Criterion(BIC) and Mean According to the results it is expected the
Absolute Percentage Error(MAPE). number of Dengue patients is increasing in
first two months in 2017 in Kurunegala
Therefore, the final model is SARIMA district.
(0,1,0) (1,1,1)6. Then by using the
coefficient table 2 REFERENCES
x = x + x - x + z -z (12)  Dengue mega project. (2015). Retrieved
t
t
t-1
t-6
t-6
t-7
from
Then the Dengue patient count of next
six months was forecasted using the final http://www.research/Dengue/DengueMega
Project-
model. INTRODUCTION&SUMMARY.htm
Table 4: Forecasted Dengue patient count  Dengue mega project. (2015). Retrieved
for next six months from http://www.
research/Dengue/DengueMegaProject-
Month Forecasted Actual count Introduction & Summary.htm
count  Epidemiology unit,Ministry of Health.
September 162 125 (n.d.). Retrieved from www.epid.gov.lk:
October 144 67 http://www.epid.gov.lk/web/index.php?opt
November 183 110
December 245 119 ion=com_casesanddeaths&Itemid=448&la
January 314 313 ng=en
February 233 374  Epidemiology unit,Ministry of Health.
(n.d.). Retrieved from www.epid.gov.lk:
http://www.epid.gov.lk/web/index.php?opt
ion=com_casesanddeaths&Itemid=448&la
ng=en

36