Forecast of canned fish consumption in Peru for an industrial fisheries project using time series models

Abstract

In the canned fish production program, it is very important to calculate its forecast through statistical models that minimize the error of the projections and that allow estimating the quantities to be produced. The objective of this research work is to select a forecast model for the consumption of canned fish in Peru for an industrial fishing project using time series models. Prediction models such as linear regression, time series decomposition and Winters’ method were used. The input data was the monthly domestic sales of canned fish from the years 2011 to 2014. The prediction error measures such as the mean absolute deviation (MAD) and the mean absolute percentage error (MAPE) of the prediction of a company were compared for a year (2014), two years (2013-2014), three years (2012-2014) and four years (2011-2014) to validate with the prediction for the years 2015-2019. The prediction model selected is the seasonal additive time series decomposition with data from two years (2013-2014) because it obtained the lowest MAD = 588.0 and the lowest MAPE = 15.00%.