IEEE Consumer Electronics Magazine - November/December 2021 - 47

Table 1. List of abbreviations.
Abbreviation
GDP
GNI
WOFOST
PBS
AR
ARIMA
RMSE
NN
LM
ANN
RNN
Table 2. Forecasting models and techniques used in the
literature.
Meaning
Gross domestic product
Gross national income
WOrld FOod STudies
Pakistan bureau of statistics
Auto regression
Auto regressive integrated moving
average
Root-mean-square error
Neural network
Linear model (regression)
Artificial neural network
Recurrent neural network
#
Model
Technique
1 WOFOST Linear Regression
2 ARIMA-SVM
SVM
3 Discovery Algo.
LSTM-NN
4
5 WARM
ARIMA
6
Multiple LR
Neural Networks
WARM + RS
Adaptive ARIMA
7 Forecast Engine Neural Networks
Citation
8
9
10
11
12
13
14
on forecasting the consumption of major crops
by not only relying on the consumption data of
previous years, but developed models by including
economic, social, and environmental factors
(described precisely in the following section)
with modified AR(1) forecasting models.
efforts in order to forecast the consumption and
production of major crops in Pakistan with varyingdegree
of success. Fongand Usman5 discuss
the machine learning for end consumers, ML and
data mining can be applied to construct predictive
pattern-based models that demonstrate statistical
power, as well as help us discover and understand
hidden trends and patterns inherent in data.
These hidden trends and patterns can be used to
formulate actionable knowledge. Actionability is
one of the most important measures of interestingness
and analysts thrive to discover knowledge
that they can act upon to their advantage.
In a study, researchers applied an energy-efficient
protocol for precision agriculture. The proposed
protocol provides efficient coverage of
sensors to the entire agriculture area.6 Kumar
et al. proposed smart sensing to facilitate quantitative
and qualitative improvement in productions.
Smart sensing for agriculture (SSA) not
only increases the monetary benefits to farmers
but also ensures delivery of fresh products.7
In the literature, a large number of studies
have been studied, focused mainly on the production
and yield forecasting of single crop at a
time. Table 2 describes plentiful work in this
area, provides a fruitful involvement to research
while gap exists to the significance of forecasting
consumption of food crops and cash crops
simultaneously. This research work is focused
November/December 2021
MATERIAL AND METHODOLOGY
Accuracy of forecasting procedures gets
affected with increase in complexity of extrapolative,
judgmental, and causal methods. Simple
evidence-based forecasting measures are much
simpler to understand, and they hold higher
accuracy. Only one forecasting model cannot be
appropriate for accurate or near forecasts, so a
number of methods can be evaluated on the
basis of available datasets and parameters.15
For the proposal of consumption forecasting
of hand-picked crops, i.e., wheat, rice, and cotton,
a long range of yearly time series data is
used in this study. Respective dataset covers the
statistics about concerned socio-economic factors
ranges from 1969-1670 to 2012-2013. The
last few years are discarded due to the incomplete
reliable statistics of parameters during the
stage of data analysis. The data for all 18 dimensions
for three primary crops has been gathered
from several statistical government publications
such as Agriculture Statistics of Pakistan issued
by Pakistan Bureau of Statistics.
Dataset Backdrop
To verify the effectiveness of our proposed
model, we chose related dataset from Pakistan
because of the following motivations: According
47
IEEE Consumer Electronics Magazine - November/December 2021

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - November/December 2021
