Decision trees in selection of featured determined food quality

Applications of machine learning techniques for enhancing nondestructive food quality and safety detection

In considering the need of people all over the world for high-quality food, there has been a recent increase in interest in the role of nondestructive and rapid detection technologies in the food industry. Moreover, the analysis of data acquired by most nondestructive technologies is complex, time-consuming, and requires highly skilled operators. Meanwhile, the general applicability of various chemometric or statistical methods is affected by noise, sample, variability, and data complexity that vary under various testing conditions. Nowadays, machine learning (ML) techniques have a wide range of applications in the food industry, especially in nondestructive technology and equipment intelligence, due to their powerful ability in handling irrelevant information, extracting feature variables, and building calibration models. The review provides an introduction and comparison of machine learning techniques, and summarizes these algorithms as traditional machine learning (TML), and deep learning (DL). Moreover, several novel nondestructive technologies, namely acoustic analysis, machine vision (MV), electronic nose (E-nose), and spectral imaging, combined with different advanced ML techniques and their applications in food quality assessment such as variety identification and classification, safety inspection and processing control, are presented. In addition to this, the existing challenges and prospects are discussed. The result of this review indicates that nondestructive testing technologies combined with state-of-the-art machine learning techniques show great potential for monitoring the quality and safety of food products and different machine learning algorithms have their characteristics and applicability scenarios. Due to the nature of feature learning, DL is one of the most promising and powerful techniques for real-time applications, which needs further research for full and wide applications in the food industry.

Characterization and Discrimination of Apples by Flash GC E-Nose: Geographical Regions and Botanical Origins Studies in China

Forty-one apple samples from 7 geographical regions and 3 botanical origins in China were investigated. A total of 29 volatile compounds have been identified by flash GC E-nose. They are 17 esters, 5 alcohols, 3 aldehydes, 1 ketone, and 3 others. A principal component analysis was employed to study the relationship between varieties and volatiles. A partial least squares discriminant analysis (PLS-DA), stepwise linear discriminant analysis (SLDA), and decision tree (DT) are used to discriminate apples from 4 geographical regions (34 apple samples) and 3 botanical origins (36 apple samples). The most influential markers identified by PLS-DA are 2-hexadecanone, methyl decanoate, tetradecanal, 1,8-cineole, hexyl 2-butenoate, (Z)-2-octenal, methyl 2-methylbutanoate, ethyl butyrate, dimethyl trisulfide, methyl formate, ethanol, S(-)2-methyl-1-butanol, ethyl acetate, pentyl acetate, butyl butanoate, butyl acetate, and ethyl octanoate. From the present work, SLDA reveals the best discrimination results in geographical regions and botanical origins, which are 88.2% and 88.9%, respectively. Although machine learning DT is attempted to classify apple samples, the results are not satisfactory.

Machine learning-assisted ammonium detection using zinc oxide/multi-walled carbon nanotube composite based impedance sensors

We report a machine learning approach to accurately correlate the impedance variations in zinc oxide/multi walled carbon nanotube nanocomposite (F-MWCNT/ZnO-NFs) to NH4+ ions concentrations. Impedance response of F-MWCNT/ZnO-NFs nanocomposites with varying ZnO:MWCNT compositions were evaluated for its sensitivity and selectivity to NH4+ ions in the presence of structurally similar analytes. A decision-making model was built, trained and tested using important features of the impedance response of F-MWCNT/ZnO-NF to varying NH4+ concentrations. Different algorithms such as kNN, random forest, neural network, Naïve Bayes and logistic regression are compared and discussed. ML analysis have led to identify the most prominent features of an impedance spectrum that can be used as the ML predictors to estimate the real concentration of NH4+ ion levels. The proposed NH4+ sensor along with the decision-making model can identify and operate at specific operating frequencies to continuously collect the most relevant information from a system.

Data analysis strategies for the characterization of chemical contaminant mixtures. Fish as a case study

Thousands of chemicals are potentially contaminating the environment and food resources, covering a wide spectrum of molecular structures, physico-chemical properties, sources, environmental behavior and toxic profiles. Beyond the description of the individual chemicals, characterizing contaminant mixtures in related matrices has become a major challenge in ecological and human health risk assessments. Continuous analytical developments, in the fields of targeted (TA) and non-targeted analysis (NTA), have resulted in ever larger sets of data on associated chemical profiles. More than ever, the implementation of advanced data analysis strategies is essential to elucidate profiles and extract new knowledge from these large data sets. Specifically focusing on the data analysis step, this review summarizes the recent progress in integrating data analysis tools into TA and NTA workflows to address the challenging characterization of chemical mixtures in environmental and food matrices. As fish matrices are relevant in both aquatic pollution and consumer exposure perspectives, fish was chosen as the main theme to illustrate this review, although the present document is equally relevant to other food and environmental matrices. The key features of TA and NTA data sets were reviewed to illustrate the challenges associated with their analysis. Advanced filtering strategies to mine NTA data sets are presented, with a particular focus on chemical filters and discriminant analysis. Further, the applications of supervised and unsupervised multivariate analysis methods to characterize exposure to chemical mixtures, and their associated challenges, is discussed.

Determinants of Farmland Abandonment in Selected Metropolitan Areas of Poland: A Spatial Analysis on the Basis of Regression Trees and Interviews with Experts

Dynamic land use changes in metropolitan areas are global phenomena. The influence of urbanisation processes on farmland is twofold: urban encroachments predominantly take place at the expense of farmland, and also result in farmland abandonment processes, especially in Central Eastern and Southern Europe. This paper analyses determinants of farmland abandonment in 280 municipalities situated in six selected Polish metropolitan areas. The analysis, which covers secondary statistical data as well as primary data collected via a survey among experts, applies the regression tree method. Within the six selected metropolitan areas nearly 9% of the farmland is permanently excluded from agricultural production (actual abandonment), plus another 11.5% is currently not being used for production (semi-abandonment). For actual abandonment, physical and economic sizes of farms, part-time farming, and soil quality constitute the most relevant determinants. Socio-economic variables play a more important role in explaining semi-abandonment than actual abandonment. Temporary exclusion of farmland from agricultural production is connected with urbanisation processes. Higher shares of built-up and urbanised areas, higher population densities, and positive migration rates result in higher shares of semi-abandonment. Naturally, areas characterised by agrarian fragmentation, where due to low agricultural incomes farmers more often decided to abandon agricultural production, were, in particular, subject to this process.

Alternative data mining/machine learning methods for the analytical evaluation of food quality and authenticity - A review

In recent years, the variety and volume of data acquired by modern analytical instruments in order to conduct a better authentication of food has dramatically increased. Several pattern recognition tools have been developed to deal with the large volume and complexity of available trial data. The most widely used methods are principal component analysis (PCA), partial least squares-discriminant analysis (PLS-DA), soft independent modelling by class analogy (SIMCA), k-nearest neighbours (kNN), parallel factor analysis (PARAFAC), and multivariate curve resolution-alternating least squares (MCR-ALS). Nevertheless, there are alternative data treatment methods, such as support vector machine (SVM), classification and regression tree (CART) and random forest (RF), that show a great potential and more advantages compared to conventional ones. In this paper, we explain the background of these methods and review and discuss the reported studies in which these three methods have been applied in the area of food quality and authenticity. In addition, we clarify the technical terminology used in this particular area of research.

A QSPR study on the liquid crystallinity of five-ring bent-core molecules using decision trees, MARS and artificial neural networks

We present an approach for the prediction of liquid crystallinity of five-ring bent-core molecules. Reported classifiers can be also used for the estimation of influence of structural modifications on LC phase formation and its stability.

Metabolic profiling by (1)H NMR of ground beef irradiated at different irradiation doses

This work describes a metabolic profiling study of non-irradiated and irradiated beef (at 2.5, 4.5 and 8 kGy) using (1)H NMR and chemometrics. The assignment of all major NMR signals of the aqueous/methanolic extracts was performed. A comprehensive multivariate data analysis proved the ability to distinguish between the irradiated and non-irradiated beef. Classification trees revealed that three metabolites (glycerol, lactic acid esters and tyramine or a p-substituted phenolic compound) are important biomarkers for classification of the irradiated and non-irradiated beef samples. Overall, the achieved metabolomic results show that the changes in the metabolic profile of meat provide a valuable insight to be used in detecting irradiated beef. The use of the NMR-based approach simplifies sample preparation and decrease the time required for analysis, compared to available official analytical procedures.