SIMPLE AND DOUBLE EXPONENTIAL SMOOTHING METHODS WITH DESIGNED INPUT DATA FOR FORECASTING A SEASONAL TIME SERIES: IN AN APPLICATION FOR LIME PRICES IN THAILAND
Keywords:
Forecasting, simple and double exponential smoothing, seasonality, holtwinters, lime pricesAbstract
In this paper, the simple exponential smoothing (SES) and double exponential smoothing(DES) methods with designed input data are presented to forecast lime prices in Thailandduring the period January 2016 to December 2016. The lime prices from January 2011 toDecember 2015 are the input data (i.e. seasonal data) which were gathered from thedatabase of Simummuang market, Thailand. The major contribution of our paper is that,although, in general, the forecasting accuracy by the traditional SES and DES methodssignificantly decreases when those methods are used to forecast the data which showseasonality patterns, the proposed solution can properly handle such a problem. For thispurpose, to forecast lime prices, 5 different input data are defined before being assigned tothe SES and the DES methods: a) the monthly data of the recent year, b) the averagemonthly data of the past years, c) the median of the monthly data of the past years, d) themonthly data of the past years after applying the linear weighting factor, and e) the averagemonthly data of the past years after applying the exponential weighting factor. Thesedesigned input data are used as agents of the raw data. Our research results indicate thatusing the DES method with input b) and the optimal initial values to forecast lime pricesduring January 2016 to September 2016 significantly gives the smallest forecasting errormeasured by the mean absolute percentage error (MAPE). The forecast lime prices ofOctober 2016 to December 2016 are also given. Additionally, we also demonstrate that, inour case, the SES and the DES methods with designed input data show a smaller MAPEthan the methods using the multiplicative Holt-Winters and the additive Holt-Wintersmodels which are designed for forecasting the seasonal data.
References
Alon, I., Qi, M., and Sadowski, R.J. (2001). Forecastingaggregate retail sales: a comparison of artificialneural networks and traditional methods. Journal ofRetailing and Consumer Services, 8(3):147-156.
Booranawong, T. and Booranawong, A. (2017). Anexponentially weighted moving average methodwith designed input data assignments forforecasting lime prices in Thailand. J. Teknol.,79(6):53-60.
Brown, R.G. (1956). Exponential smoothing for predictingdemand. Proceedings of the 10th National Meetingof the Operations Research Society of America;November 16, 1956; San Francisco, CA, USA, p. 1-15.
Chatfield, C. (2001). Time-series Forecasting. Chapman &Hall, NY, USA, 280p.
De Gooijer, J.G. and Hyndman, R.J. (2006). 25 years oftime series forecasting. Int. J. Forecasting,22(3):443-473.
Dhakre, D.S., Sarkar, K.A., and Manna, S. (2016).Forecasting price of Brinjal by Holt Wintersmethod in west Bengal using MS Excel. Inter. J.Bio-res., Environ. Agri. Sci., 2(1):232-235.
Gardner, E.S. (1985). Exponential smoothing: the state ofthe art. J. Forecasting, 4:1-28.
Gentry, T.W., Wiliamowski, B.M., and Weatherford, L.R.(1995). A comparison of traditional forecastingtechniques and neural networks. IntelligentEngineering Systems through Artificial NeuralNetworks, 5:765-770.
Holt, C.C. (2004). Forecasting seasonals and trends byexponentially weighted moving averages. Int. J.Forecasting, 20(1):5-10.
Hunter, J.H. (1986). The exponentially weighted movingaverage. J. Qual. Technol., 18(4):203-210.
Kaewsuksaeng, S., Tatmala, N., Srilaong, V., andPongprasert, N. (2015). Postharvest heat treatmentdelays chlorophyll degradation and maintainsquality in Thai lime (Citrus aurantifolia Swingle cv.Paan) fruit. Postharvest Biol. Tec., 100:1-7.
Kalekar, P.S. (2004). Time series forecasting using Holt-Winters exponential smoothing. Kanwal RekhiSchool of Information Technology, Mumbai, India,13p.
Montgomery, D.C., Jennings, C.L., and Kulahci, M.(2008). Introduction to Time Series Analysis andForecasting. John Wiley and Sons, Hoboken, NJ,USA, 472p.
Pranamornkith, T. (2009). Effects of postharvesttreatments on storage quality of lime (Citruslatifolia Tanaka) fruit, [Ph.D. thesis]. MasseyUniversity, Wellington, New Zealand, 261p.
Ramos, P., Santos, N., and Rebelo, R. (2015). Performanceof state space and ARIMA models for consumerretail sales forecasting. Robot. Cim-Int. Manuf.,34:151-163.
Ravindran, A. and Warsing, D.P. (2013). Supply chainengineering: models and applications. CRC Press,NY, USA, 548p.
Tratar, L.F. (2016). Forecasting method for noisy demand.Int. J. Prod. Econ., 161:64-73.
Tratar, L.F., Mojskerc, B., and Toman, A. (2016). Demandforecasting with four-parameter exponentialsmoothing. Int. J. Prod. Econ., 181:162-173.
Tratar, L.F. and Srmcnik, E. (2016). The comparison ofHolt-Winters method and regression methods: acase study. Energy, 109:266-276.
Winters, P. (1960). Forecasting sales by exponentiallyweighted moving averages. Manage. Sci., 6:324-342.
