Page 90 - Contributed Paper Session (CPS) - Volume 1

CPS1119 Fethi Şaban ÖZBEK
The purpose of this paper is to evaluate the effectiveness of artificial
neural network as a forecasting tool, and to forecast the agricultural
commodity prices in Turkey for the period from October 2018 to December
2019.

2. Methodology
ANN, represents a nonlinear statistical modelling tool that is based on the
concept of a biological neural network (Shahinfar et al., 2012), was used for
forecasting prices of wheat, barley, maize (seed), and raw cotton in the
agriculture of Turkey. The data of monthly price from 2009 January to 2018
September were used as sample data (network training) and historical
prediction evaluation (network evaluation). In each forecasting, the input was
the last observation, and the output was the predicted value of agricultural
commodity price.
Different ANN models for different problem structures have been
developed (Akkol et al., 2017). The most used networks in literature are
known as single and multilayer perception, vector quantization models
(LVQ), self-organizing model (SOM), adaptive resonance theory models (ART),
Hopfield networks, Elman network and radial based networks (Öztemel,
2006). In this study, the multilayer perception model of ANN was used. The
multilayer perception network is a function of predictors (also called inputs or
independent variables) that minimize the prediction error of target variables
(also called outputs). This structure is known as feedforwardarchitecture
because the acquaintances in the network flow forward from the input layer
to the output layer without any feedback loops (Giri et al., 2014).
Node in the hidden layer contains hyperbolic tangent activation function
(Eq. 1). It takes real-valued arguments and transforms them to the range (–1,
1).
y
e i−e −y i
f(y ) = e i+e −y i Eq. 1
i
y
And they take a weighted sum of all input variables:

y = ∑ w x Eq. 2
i
j
ji i
where x is an input variable and w is corresponding weight in layer j.
i
ji

Identity function was used in output layer activation. Identity function has the
form in Eq. 3. It takes real-valued arguments and returns them unchanged
(IBM SPSS Neural Networks 22, 2018).

f(y ) = y Eq. 3
i
i

79 | I S I W S C 2 0 1 9