Browsing by Author "Peiris, T. S. G"

Now showing 1 - 3 of 3

Open Access
Development of SARIMA Model to Predict Quarterly Apparel and Textile Export Revenue in Sri Lanka
(Faculty of Humanities and Sciences, SLIIT, 2023-11-01) Piyasiri, K. G. V.; Kasthuriarachchi, U. P; Nirmani, K. G. R; Tilakaratne, K.I.; Peiris, T. S. G
Apparel and textile exports play a significant role in the Sri Lankan economy. The USA, UK, Italy, Germany, and Belgium are the main markets of apparel and textile exports in Sri Lanka. Advanced knowledge of export revenue is vital important for various reasons. A Seasonal Autoregressive Integrated Moving Average (SARIMA) model of the type (1,1,0) x (0,1,1)4 was developed to model apparel and textile export revenue in Sri Lanka using quarterly data from year 2004 quarter 1 (2004Q1) to year 2021 quarter 4 (2021Q4). The errors of the model were found to be random and have a constant variance. The best fitted model was identified by comparing various statistical indicators, namely, the Akaike info criterion, Schwarz criterion, Hannan-Quinn criterion, Log likelihood criterion and volatility of six possible models decided based on sample ACF and PACF of the stationary series. The model was validated for data from year 2022Q1 to 2023Q1. The Mean Absolute Percentage Error (MAPE) for the training data set and validation data set were 7.68% and 11.35% respectively. The predicted revenues (Mn USD) for the 2023Q2 to 2024Q4 are 1074.23, 1263.30, 1222.22, 1206.74, 1058.38, 1265.00 and 1216.58, respectively. The forecasted values for short-term periods can be effectively used by the decision makers for various activities. The model developed is easy to use and reliable.
Open Access
Forecasting of Constant GDP per capita of Sri Lanka using ARIMA model
(Faculty of Humanities and Sciences, SLIIT, 2023-11-01) Wijesinghe, W.R.A.N.M; Umayangi, K.A.S,; De Silva, H.D.K; Mundigalage, S.D.M; Peiris, T. S. G
GDP per capita is a global measurement for assessing the economic prosperity of nations. Constant (Real) GDP per capita eliminates the effects of inflation which allows for a more accurate comparison of GDP per capita over time. However, no statistical models have been developed to predict annual constant GDP per capita (CGDPC) in Sri Lanka. In this study, ARIMA (1,1,0) model was developed using past data from 1961 to 2018 to forecast CGDPC. The best-fitted model was identified based on three possible models using sample ACF and sample PACF of the stationary series and comparing statistics such as AIC, BIC, maximum log-likelihood, and volatility. The residuals of the fitted model were white noise. The training dataset has percentage errors ranging from -6.50% to 3.80%. The model was validated for observed data in 2019, 2020, and 2021. The percentage error for the three points were -3.49, -6.10, and 1.49 respectively. The forecasted values for 2022, 2023, and 2024 obtained were 4506.728, 4653.895, and 4810.505 respectively showing that Sri Lanka’s economy is expected to grow due to the increase in CGDPC. The GDP per capita growth rates of 2.99%, 3.27%, and 3.37% for the next 3 years also confirm this. The results obtained from this study can be effectively used for better planning. However, it is recommended to improve the model further to reduce the percentage of errors using the ARIMAX approach.
Embargo
Modeling and Forecasting of the Weekly Incidence of Dengue in Colombo District of Sri Lanka
(Faculty of Humanities and Sciences, SLIIT, 2022-09-15) Arachchi, K. A. N. L. K.; Peiris, T. S. G
This study was designed to develop a time series model for the weekly incidence of dengue in the Colombo district of Sri Lanka. Weekly occurrence of dengue fever counts from January 2015 to August 2020 in the Epidemiological Report by the Ministry of Health was used for the study . ARIMA (2,1,0) with the addition of AR (16) was identified as the most effective model. The model was trained using data from January 2015 to December 2019. The balance data was used to validate the model. The residuals of the model satisfied the randomness and constant variance, but the residuals significantly deviated from the normality. The results showed that the forecasted figures were consistent with the observed series. However, a noticeable percentage error was observed sequentially in the late 2020s. Those errors could be attributable to the fact that there was an underreporting of dengue fever cases due to social and operational shocks of the Covid-19 Pandemic.