CFE 2015: Start Registration
View Submission - CFE
A1418
Topic: Contributions on forecasting Title: Short-term salmon price forecasting Authors:  Daumantas Bloznelis - Norwegian University of Life Sciences (Norway) [presenting]
Abstract: Price forecasting has a fundamental value in the highly volatile market of farmed Atlantic salmon. Accurate price predictions could reduce uncertainty and aid planning decisions of salmon farmers, processors and other market participants. We forecast salmon spot prices by univariate and multivariate time series models, artificial neural networks and the $k-$nearest neighbours method using weekly data for 2007-2014. A naive seasonal forecast reduces the mean absolute error of a naive no-change forecast by up to 5$\%$, and predicts the direction of the price change correctly up to 65$\%$ of the time for 1-5 weeks ahead. However, no method seems to consistently improve upon the naive seasonal forecast, lending support to a hypothesis of weak form efficiency of the salmon spot market.