-*- mode: org -*-
Food industry
*  33(<-122): Assuring the authenticity of northwest Spain white wine varieties using machine learning techniques

Classification of wine represents a multi-criteria decision-making
problem characterized by great complexity, non-linearity and lack of
objective information regarding the quality of the desired final
product. Volatile compounds of wines elaborated from four Galician (NW
Spain) autochthonous white Vitis vinifera from four consecutive
vintages were analysed by gas chromatography (F1D, FPD and MS
detectors), and several aroma compounds were used for correctly
classifying autochthonous white grape varieties (Albarifio,
Treixadura, Loureira and Dona Branca). The objective of the work is
twofold: to find a classification model able to precisely
differentiate between existing grape varieties (i.e. assuring the
authenticity), and to assess the discriminatory power of different
family compounds over well-known classifiers (i.e. guaranteeing the
typicality). From the experiments carried out, and given the fact that
Principal Component Analysis (PCA) was not able to accurately separate
all the wine varieties, this work investigates the suitability of
applying different machine learning (ML) techniques (i.e.: Support
Vector Machines, Random Forests, MultiLayer Perceptron, k-Nearest
Neighbour and Naive Bayes) for classification purposes. Perfect
classification accuracy is obtained by the Random Forest algorithm,
whilst the other alternatives achieved promising results using only
part of the available information. (C) 2013 Elsevier Ltd. All rights
reserved.

2014


*  34(<-241): An alternative approach for the classification of orange varieties based on near infrared spectroscopy

A multivariate technique and feasibility of using near infrared
spectroscopy (NIRS) for non-destructive discriminating Thai orange
varieties were studied in this paper. Short-wavelength near infrared
(SWNIR) spectra in region of 643 to 970 nm were collected from 100
orange sample of each varieties. A total of 300 spectra were used to
develop an accurate classification model by diversity of
classifiers. The result showed that Logistic Regression (LGR) model
was achieved 100% classification accuracy while Multi-Criteria
Quadratic Programming (MCQP) and Support Vector Machine (SVM) ones
also demonstrated satisfying result (95%). In order to find simpler
and easier interpretable classification model, several feature
selection techniques were evaluated to identify the most relevant
wavelengths to the orange varieties. With four principal components
(PCs) from Principal Component Analysis (PCA) and the effective
wavelengths of 769.68, 692.28, 662.61 and 959.31 nm from Least Square
Forward Selection (LS-FS), the reduced classification models of LGR
also achieved satisfying classification accuracy
respectively. Furthermore, both Kernel Principal Component Analysis
(KPCA) and Kernel Least Square Forward Selection (KLS-FS) with SVM
enhanced performance of models by 5 PCs and features respectively. The
result concluded that NIRS can yield an accurate classification for
Thai tangerine varieties from whole spectra and can enhance
interpretability of classification model by feature subset. (C) 2013
Elsevier B.V. All rights reserved.

2013


*  35(<-370): Analysis of complex, processed substances with the use of NIR spectroscopy and chemometrics: Classification and prediction of properties - The potato crisps example

An NIR spectroscopic method was researched and developed for the
analysis of potato crisps (chips) chosen as an example of a common,
cheap but complex product. Four similar types of the 'original
flavour' potato chips from different manufacturers were analysed by
NIR spectroscopy; as well, the quality parameters - fat, moisture,
acid and peroxide values of the extracted oil were
predicted. Principal component analysis (PCA) of the NIR data
displayed the clustering of objects with respect to the type of
chips. NIR spectra were rank-ordered with the use of the sparingly
applied multiple criteria decision making (MCDM) ranking methods,
PROMETHEE (Preference Ranking Organization METHod for Enrichment
Evaluation) and GAIA (Geometrical Analysis for Interactive Aid), and a
comprehensive quantitative description of the data was obtained. The
four traditional parameters were predicted on the basis of the NIR
spectra; the performance of the Partial Least Squares (PLS), and
Kernel Partial Least Squares (KPLS) calibrations was compared with
those from Least Squares-Support Vector Machines (LS-SVM) method. The
LS-SVM calibrations, which model better data linearity and
non-linearity, successfully predicted all four parameters. This work
has demonstrated that NIR methodology with the use of chemometrics can
describe comprehensively qualitative and quantitative properties of
complex, processed substances as illustrated by the potato chips
example, and indicated that this approach may be applied to other
similar complex samples. (C) 2010 Elsevier B.V. All rights reserved.

2011



*  50(<-289): Enhancing electronic nose performance: A novel feature selection approach using dynamic social impact theory and moving window time slicing for classification of Kangra orthodox black tea (Camellia sinensis (L.) O. Kuntze)

This paper presents a novel multiobjective wrapper approach using
dynamic social impact theory based optimizer (SITU) and moving window
time slicing (MWTS) for the performance enhancement of an electronic
nose (EN). SITU, in conjunction with principal component analysis
(PCA) and support vector machines (SVMs) classifier, has been used for
the classification of samples collected from the single batch
production of Kangra orthodox black tea (Camllia sinensis (L)
U. Kuntze). The work employs a novel SITU assisted MWTS (SITO-MWTS)
technique for identifying the optimum time intervals of the EN sensor
array response, which give the maximum classification rate. Results
show that, by identifying the optimum time slicing window positions
for each sensor response, the performance of an EN can be
improved. Also, the sensor response variability is time dependent in a
sniffing cycle, and hence good classification can be obtained by
selecting different time intervals for different sensors. The proposed
method has also been compared with other established techniques for EN
feature extraction. The work not only demonstrates the efficacy of
SITU for feature selection owing to its simplicity in terms of few
control parameters, but also the capability of an EN to differentiate
Kangra orthodox black tea samples at different production stages. (C)
2012 Elsevier B.V. All rights reserved.

2012





*  51(<-329): A novel approach using Dynamic Social Impact Theory for optimization of impedance-Tongue (iTongue)

This paper presents a novel multiobjective wrapper approach using
Dynamic Social Impact Theory based optimizer (SITO). A Fuzzy Inference
System in conjunction with support vector machines classifier has been
used for the optimization of an impedance-Tongue for the
classification of samples collected from single batch production of
Kangra orthodox black tea. Impedance spectra of the tea samples have
been measured in the range of 20 Hz to 1 MHz using a two electrode
setup employing platinum and gold electrodes. The proposed approach
has been compared, for its robustness and validity using various intra
and inter measures, against Genetic Algorithm and binary Particle
Swarm Optimization. Feature subset selection methods based on the
first and second order statistics have also been employed for
comparisons. The proposed approach outperforms the Genetic Algorithm
and binary Particle Swarm Optimization. (C) 2011 Elsevier B.V. All
rights reserved.

2011





* 376(<-328): Food processing optimization using evolutionary algorithms

Evolutionary algorithms are widely used in single and multi-objective
optimization. They are easy to use and provide solution(s) in one
simulation run. They are used in food processing industries for
decision making. Food processing presents constrained and
unconstrained optimization problems. This paper reviews the
development of evolutionary algorithm techniques as used in the food
processing industries. Some evolutionary algorithms like genetic
algorithm, differential evolution, artificial neural networks and
fuzzy logic were studied with reference to their applications in food
processing. Several processes involved in food processing which
include thermal processing, food quality, process design, drying,
fermentation and hydrogenation processes are discussed with reference
to evolutionary optimization techniques. We compared the performances
of different types of evolutionary algorithm techniques and suggested
further areas of application of the techniques in food processing
optimization.

2011




* 603(<-191): Modeling and Optimization of Electrodialytic Desalination of Fish Sauce Using Artificial Neural Networks and Genetic Algorithm

Electrodialysis (ED) has been proposed as a means to reduce sodium ion
concentration in fish sauce. However, no information is so far
available on the optimum condition to operate the ED
process. Artificial neural network (ANN)-based models were therefore
developed to predict the ED performance and changes in selected
quality attributes of ED-treated fish sauce; optimum operating
condition of the process was then determined via multi-objective
optimization using genetic algorithm (MOGA). The optimal ANN models
were able to predict the ED performance with R (2) = 0.995, fish sauce
basic characteristics with R (2) = 0.992, and the concentrations of
total aroma compounds and total amino acids, flavor difference, and
saltiness of the treated fish sauce with R (2) = 0.999. Through the
use of MOGA, the optimum condition of the ED process was the use of an
applied voltage of 6.3 V and the maintenance of the residual salt
concentration of the treated fish sauce of 14.3 % (w/w).

2013




* 662(<-117): Optimization of response surface and neural network models in conjugation with desirability function for estimation of nutritional needs of methionine, lysine, and threonine in broiler chickens

The optimization algorithm of a model may have significant effects on
the final optimal values of nutrient requirements in poultry
enterprises. In poultry nutrition, the optimal values of dietary
essential nutrients are very important for feed formulation to
optimize profit through minimizing feed cost and maximizing bird
performance. This study was conducted to introduce a novel
multi-objective algorithm, desirability function, for optimization the
bird response models based on response surface methodology (RSM) and
artificial neural network (ANN). The growth databases on the central
composite design (CCD) were used to construct the RSM and ANN models
and optimal values for 3 essential amino acids including lysine,
methionine, and threonine in broiler chicks have been reevaluated
using the desirable function in both analytical approaches from 3 to
16 d of age. Multi-objective optimization results showed that the most
desirable function was obtained for ANN-based model (D - 0.99) where
the optimal levels of digestible lysine (dLys), digestible methionine
(dMet), and digestible threonine (dThr) for maximum desirability were
13.2, 5.0, and 8.3 g/kg of diet, respectively. However, the optimal
levels of dLys, dMet, and dThr in the RSM-based model were estimated
at 11.2, 5.4, and 7.6 g/kg of diet, respectively. This research
documented that the application of ANN in the broiler chicken model
along with a multi-objective optimization algorithm such as
desirability function could be a useful tool for optimization of
dietary amino acids in fractional factorial experiments, in which the
use of the global desirability function may be able to overcome the
underestimations of dietary amino acids resulting from the RSM model.

2014





* 667(<-485): Baking of Flat Bread in an Impingement Oven: Modeling and Optimization

An artificial neural network (ANN) was developed to model the effect
of baking parameters on the quality attributes of flat bread; i.e.,
crumb temperature, moisture content, surface color change and bread
volume increase during baking process. As the hot air impinging jets
were employed for baking, the baking control parameters were the jet
temperature, the jet velocity, and the time elapsed from the beginning
of the baking. The data used in the training of the network were
acquired experimentally. In addition, using the data provided by ANN,
a multi-objective optimization algorithm was employed to achieve the
baking condition that provides the quality of the bread in all aspects
simultaneously.

2009