Tree Network Models Research Articles

Development of Spectroscopic Sensor System for an IoT Application of Adulteration Identification on Milk Using Machine Learning

Adulteration in milk is a common scenario for gaining extra profit, which may cause severe harmful effects on humans. The qualitative spectroscopic technique provides a better solution for detecting the toxic contents of milk and foodstuffs. All the available spectroscopic methods for milk adulterant detection are based on laboratory-based with costly equipment. This laboratory-based detection takes a long time and is more expensive, which may not be afforded by a common man. To overcome this issue, this research work involves the design and development of a low-cost, portable, multispectral, AI-based, non-destructive spectroscopic sensor system that can be used to detect the milk adulterant in real-time. The designed sensor system uses the spectroscopic method with wavelength ranges from (410-940nm) which consists of three different bands Ultraviolet (UV), visible, and Infra-Red(IR) spectrum to improve the accuracy of detection. The sensor system is connected to the internet via the developed IoT application module, which displays the detected adulterant results in a dedicated web page designed for this purpose. This IoT application enables the adulterant detected results published on the internet immediately with location information for bringing transparency. Adulterant detection problem is formulated as a classification problem and solved by machine learning algorithms of a decision tree, Naive Bayes, linear discriminant analysis, support vector machine and neural network model. The average accuracy of linear discriminant analysis, support vector machine, Naive Bayes, decision tree and neural network model are obtained as 88.1%, 90%, 90%, 91.7% and 92.7% respectively. Genetic algorithm framework is formulated for hyperparameter tuning of neural network model which improved the accuracy from 92.7% to 100%. The model is trained for five different classes of four adulterants, namely Sodium Salicylate, Dextrose, Hydrogen Peroxide, Ammonium Sulphate, and one pure milk sample.

Machine Learning Techniques to Predict Slope Failures in Uttarkashi, Uttarakhand (India)

Uttarkashi region is highly prone to landslides because of its geological structure. The exact occurrence of these landslides events is difficult to predict due to its complex mechanism and dependence on the number of triggering factors. Moreover, the behavior and prediction of unstable slopes are of high importance failing of which otherwise can have a devastating impact. This research work aims at studying and modeling slopes with the help of state-of-the-art machine learning techniques: Support vector machine, Backpropagation, Decision tree, and Bayesian Network models. To train and test these models a total of 137 instances are taken. Moreover, the independence of individual features is studied with the help of multicollinearity analysis. The capability of considered methods was evaluated using various performance evaluation metrics. Evaluation and comparison of the results show that the performance of all classifiers is satisfactory for slope failure analysis (AUC=0.750-0.889). Based on the results Bayesian network proved to be most efficient to predict slope failures (Accuracy=93%, AUC=.889). These outcomes can be beneficial for government agencies in early-stage risk mitigation.

A novel approach for classification of soils based on laboratory tests using Adaboost, Tree and ANN modeling

This research focuses on presenting new models based on classifiers that can be applied to various problems. Adaboost is a type of ensemble learning machine that uses classifiers that contain a range of base models. This study used enhanced Adaboost models to classify soil types base on tree algorithm models that are less commonly used in this area. Determining the type of soil in different geotechnical projects is very important. Using soil classification, soil properties such as mechanical properties, performance against static and dynamic loads can be found. Regarding the importance of the subject, 440 samples of the actual project were used to design this new methodology. The dataset included clay content, moisture content, specific gravity, void ratio, plastic, and liquid limit parameters to determine the type of soil classification. These samples were tested with high precision and the actual type of classification was obtained. For comparison, two enhanced tree and neural network model were designed and developed according to these conditions. The results of this classification were presented for different soil samples. The developed adaboost model showed that it could well classify the soil. This model showed that only 11 samples were not correctly identified among the total data (88 data). Therefore, this new technique can be used to increase the accuracy and reduce the cost of projects.

Assessment of the State of Quality in Garments Applying Data Mining Mechanisms: A Case Study in the Apparel Industry

Forecasting the quality of sewed garments is an important area in the apparel industry. This paper consists of a case study relevant to a high-ranking apparel manufacturing plant in Sri Lanka. Quality is measured using the First Time Through (FTT) state which is a measure of production competence and capacity. The factory capacity is to afford the FTT 98% or above as a high state category. The low state consists of FTT of less than 98%. Recently Data mining methods are used to extract insights from data and to make fast decisions. The main objective of the study is to identify a better model to predict the FTT state with data mining mechanisms. Classification tree and Probabilistic Neural Network (PNN) models were used to forecast the FTT state with the under-sampling method due to the matter of class imbalance in the original dataset. True positive (TP), False-positive (FP), precision, recall, accuracy and F-measure were used as the performance measurements. FP rate was zero and precision was one in the classification tree. While the FP rate was 0.0649 and precision was 0.9348 in the PNN model. Both models had a high F-measure value of 0.9745 and 0.9287 respectively. Therefore, two models can be used in prediction with better performance measurements. Outcomes of the study will help to find out the optimum allocation of a style to a relevant team to achieve the highest FTT state, to recognize the training requirements of the employees and to improve the satisfaction of the customer.

Exploratory, Regression, and Neural Network Analysis of the Stability of Cation Coronates in Selected Pure Solvents

Exploratory, regression, and neural network analysis of the stability constants of crown ether [12C4, 16C5, (CH3)216C5, DB21C7, DB24C8, DCH24C8, DB30C10] 1 : 1 complexes with alkaline (Li+, Na+, K+, Cs+, Rb+), alkaline-earth (Ca2+, Sr2+, Ba2+), and heavy (Ag+, Tl+, Co2+, Cu2+, Pb2+) metals and NH4+ in water and organic solvents (methanol, acetonitrile, acetone, N,N-dimethylformamide, nitrobenzene, nitromethane, 1,2-dichloroethane, propylene carbonate) at 298.15 K obtained via conductometry has been performed. Factor, cluster, discriminant, canonical, decision tree, regression, and neural network models of clustering, approximation, and prediction of thermodynamic constants of the complexation depending on the properties of the ligand, the cation, and the solvent have been developed. The trained MLP 7-5-5 Multilayer Perceptron Cluster has completely confirmed the k-means clustering. Independent data on the stability constants of coronates have demonstrated the predictive capacity of the trained perceptron-approximator MLP 7-7-1.

EVALUATING THE EFFECT OF APERTURE VARIATION ON THE HYDRAULIC PROPERTIES OF THE THREE-DIMENSIONAL FRACTAL-LIKE TREE NETWORK MODEL

Hydraulic properties of rock fractures are important issues for many geoengineering practices. Previous studies have revealed that natural rock fractures have variable apertures that could significantly influence the permeability of single rock fractures, yet the effect of aperture variation on the hydraulic properties of fractured rock masses has received little attention. The present study implemented a series of flow calculations on two kinds of 3D fractal-like tree network models constituted by fractures with heterogeneous aperture distributions and parallel plates with uniform apertures to gain insight into the effect of aperture variation on the hydraulic properties of fractured rock masses. The fluid flows through the two kinds of models are simulated using a developed numerical code, whose validity is verified by comparisons with the analytical solutions and previous experimental results. The effects of heterogeneous aperture distributions on the permeability of the models are illustrated by changing the geometrical parameters including the aperture ratio and mean aperture of the initial fracture branch. The results show that a significant channeling flow is observed in the 3D fractal-like tree network models constituted by fractures with heterogeneous aperture distributions. These preferential flow paths are mainly located in the large aperture channels, bypassing the contacts and/or low aperture barriers. The equivalent permeability of the 3D fractal-like tree network models increases with increasing fracture aperture ratio and/or increasing mean aperture of the initial fracture branch. The permeability ratio of the model of fractures with heterogeneous aperture distributions to the model with uniform mean apertures is always smaller than 1.0, and the ratio increases following a logarithmic relationship as the mean aperture of the initial fracture branch increases. A regression function is proposed to estimate the equivalent permeability of the model of fractures with heterogeneous aperture distributions. The deviations between the predicted permeability and the simulated/experimental results are less than half of the order of magnitude, which verifies that the proposed equation is capable of predicting the equivalent permeability of 3D fractal-like tree networks as a first-order estimation.

Financial Distress Prediction and Feature Selection in Multiple Periods by Lassoing Unconstrained Distributed Lag Non-linear Models

In this paper, we propose a new framework of a financial early warning system through combining the unconstrained distributed lag model (DLM) and widely used financial distress prediction models such as the logistic model and the support vector machine (SVM) for the purpose of improving the performance of an early warning system for listed companies in China. We introduce simultaneously the 3~5-period-lagged financial ratios and macroeconomic factors in the consecutive time windows t − 3, t − 4 and t − 5 to the prediction models; thus, the influence of the early continued changes within and outside the company on its financial condition is detected. Further, by introducing lasso penalty into the logistic-distributed lag and SVM-distributed lag frameworks, we implement feature selection and exclude the potentially redundant factors, considering that an original long list of accounting ratios is used in the financial distress prediction context. We conduct a series of comparison analyses to test the predicting performance of the models proposed by this study. The results show that our models outperform logistic, SVM, decision tree and neural network (NN) models in a single time window, which implies that the models incorporating indicator data in multiple time windows convey more information in terms of financial distress prediction when compared with the existing singe time window models.

Imputation of Missing Values Affecting the Software Performance of Component-based Robots

Intelligent robots are foreseen as a technology that would be soon present in most public and private environments. In order to increase the trust of humans, robotic systems must be reliable while both response and down times are minimized. In keeping with this idea, present paper proposes the application of machine learning (regression models more precisely) to preprocess data in order to improve the detection of failures. Such failures deeply affect the performance of the software components embedded in human-interacting robots. To address one of the most common problems of real-life datasets (missing values), some traditional (such as linear regression) as well as innovative (decision tree and neural network) models are applied. The aim is to impute missing values with minimum error in order to improve the quality of data and consequently maximize the failure-detection rate. Experiments are run on a public and up-to-date dataset and the obtained results support the viability of the proposed models.

Performance evaluation of optimized and adaptive neuro fuzzy inference system for predictive modeling in agriculture

The Neural Network has a significant impact in developing predictive models in a wide range of applications. In this paper, a neuro-fuzzy prediction model is developed depending on improving the performance of the traditional artificial neural networks using Adaptive Momentum Optimizer. This optimizer simulates the behavior of the International Trade Analysis in the agriculture industry, and this method is used to determine the optimal parameters of artificial neural networks. The proposed model is compared with the existing models such as Support Vector Machine, Random Forest, Decision Tree and traditional Artificial Neural Network models. To examine the forecasting performance of the proposed approach, agriculture datasets is used. The performance of the models was assessed using different performance evaluation criteria and the empirical results show that the back propagation neural network with Adam optimizer attains favorable prediction accuracy of 96.78%, and a better convergence rate. Compared to other benchmark algorithms, the proposed algorithm performs better, and the result validates the effectiveness of the back propagation with Adam optimizer for Natural Language Processing.

ANALYTICAL SOLUTIONS FOR PERMEABILITY OF A THREE-DIMENSIONAL FRACTAL-LIKE TREE NETWORK MODEL WITH FRACTURES HAVING VARIABLE WIDTHS

This study proposed analytical solutions for permeability of a fractal-like tree network model with fractures having variable widths, which has not been reported before, if any. This model is more realistic with natural fracture networks than the traditional constant width fracture network models. The results show that considering fracture width variations decreases the permeability. Taking the fracture width ratio that equals to 0.6 and the total number of branching levels that equals to 30 as an example, the permeability decreases by more than three orders of magnitude with respect to that of a constant width fracture network model. The fracture length ratio plays a more significant role in permeability when it is larger than 0.8 than that is less than 0.8. The permeability is more sensitive to the fracture aperture ratio that is less than 0.8. When the total number of branching levels is large (i.e. 30), the permeability changes significantly (i.e. more than three orders of magnitude); whereas when the total number of branching levels is small (i.e. 5), the permeability varies in a small range (i.e. less than one order of magnitude). When taking into account the relationships among fracture length ratio, fracture aperture ratio and fracture width ratio, the parameters can be easily obtained and analytical solutions for permeability can also be easily derived. The empirical function for predicting critical hydraulic gradient is proposed, which can be used to estimate whether the fluid flow is within the linear flow regime and whether the proposed analytical solutions are applicable in the present study.

Tree networks to assess financial contagion

We propose a two-layered tree network model that decomposes financial contagion into a global component, composed of inter-country contagion effects, and a local component, made up of inter-institutional contagion channels. The model is effectively applied to a database containing time series of daily CDS spreads of major European financial institutions (banks and insurance companies), and reveals the importance of monitoring both channels to assess financial contagion. Our empirical application reveals evidence of a high inter-country and inter-institutional vulnerability at the onset of the global financial crisis in 2008 and during the sovereign crisis in 2011. The results identify France as central to the inter-country contagion in the Euro area during the financial crisis, while Italy dominates during the sovereign crisis. The application of the model to detect contagion between sectors of the European economy reveals similar findings, and identifies the manufacturing sector as the most central, while, at the company level, financial institutions dominate during the 2008 crisis.

Assessing the performance of decision tree and neural network models in mapping soil properties

To build any spatial soil database, a set of environmental data including digital elevation model (DEM) and satellite images beside geomorphic landscape description are essentials. Such a database, integrates field observations and laboratory analyses data with the results obtained from qualitative and quantitative models. So far, various techniques have been developed for soil data processing. The performance of Artificial Neural Network (ANN) and Decision Tree (DT) models was compared to map out some soil attributes in Alborz Province, Iran. Terrain attributes derived from a DEM along with Landsat 8 ETM+, geomorphology map, and the routine laboratory analyses of the studied area were used as input data. The relationships between soil properties (including sand, silt, clay, electrical conductivity, organic carbon, and carbonates) and the environmental variables were assessed using the Pearson Correlation Coefficient and Principle Components Analysis. Slope, elevation, geomforms, carbonate index, stream network, wetness index, and the band’s number 2, 3, 4, and 5 were the most significantly correlated variables. ANN and DT did not show the same accuracy in predicting all parameters. The DT model showed higher performances in estimating sand (R2=0.73), silt (R2=0.70), clay (R2=0.72), organic carbon (R2=0.71), and carbonates (R2=0.70). While the ANN model only showed higher performance in predicting soil electrical conductivity (R2=0.95). The results showed that determination the best model to use, is dependent upon the relation between the considered soil properties with the environmental variables. However, the DT model showed more reasonable results than the ANN model in this study. The results showed that before using a certain model to predict variability of all soil parameters, it would be better to evaluate the efficiency of all possible models for choosing the best fitted model for each property. In other words, most of the developed models are site-specific and may not be applicable to use for predicting other soil properties or other area.

Intelligent employment rate prediction model based on a neural computing framework and human–computer interaction platform

An intelligent employment rate prediction model based on a neural computing framework and human–computer interaction platform is demonstrated in this manuscript. Predictive analytics is the future of things, and its significance is manifested in two main aspects: understanding the future so that people can prepare for its arrival, and predicting the current decision so that people can understand the possible consequences, and by the consequences of the analysis to determine the current decision, and strive to make the current decision. However, there are lots of challenges for the prediction tasks. The novelty of this research is mainly concentrated on two major aspects: (1) the neural network model is optimized and enhanced. The proposed nerve tree network model is essentially based on a tree-structured code for a multi-layered feed-forward sparse neural network; with the tree-structured code, the nerve tree network model does not require interconversion between its genotype and phenotype in the coding and decoding operations, and also effectively reduces the computing time. (2) The human–computer interaction is integrated to construct a user-friendly system. In interactive technology, the interactive contact surface and the model, interactive methods and social acceptance have also given rise to many questions that must be solved and problems that require further research and technological innovation. Through numerical verification, the performance of the proposed framework is validated, and the simulation proves the overall performance of the proposed model. Compared with other models, the proposed algorithms can achieve higher prediction accuracy.

Predictive Models for Classifying the Outcomes of Violence: Case Study for Thailand’s Deep South

Violence is now widely recognized as a public health problem because of its significant consequences on the health and wellness of people and it remains a growing problem in many countries including Thailand. Elucidating the factors related to violence can provide information that can help to prevent violence and decrease the number of injuries. This study explored predictive data mining models which have high interpretability and prediction accuracy in classifying the outcomes of violence. After data preprocessing, a set of 21,424 incidents occurring from 2004 to 2016 were obtained from the Deep South Coordination Centre database. A correlation-based feature subset selection and decision tree technique with embedded feature selection were used for variable selection and four data mining techniques were applied to classify the violent outcomes into physical injury and no physical injury. The findings revealed that regardless of the variable selection method, gun was selected as a risk factor of physical injury. Moreover, a decision tree model with three variables, gun, zone, and solid/sharp weapon outperformed a naive Bayes model in terms of accurate performance and interpretability. Decision tree and artificial neural network models have similar levels of performance in classifying the outcome of violence but in practical terms, a decision tree model is more interpretable than an artificial neural network model.

Tree Networks to Assess Financial Contagion

We proposes a two-layered tree network model that decomposes financial contagion into a global component, composed of inter-country contagion effects, and a local component, made up of inter-institutional contagion channels. The model is effectively applied to a database containing time series of daily CDS spreads of major European financial institutions (banks and insurance companies), and reveals the importance monitoring both channels to assess financial contagion. The empirical application revealed evidence of a high inter-country and inter-institutional vulnerability at the onset of the global financial crisis in 2008 and during the sovereign crisis in 2011. The result further identifies Belgium and France as central to the inter-country contagion in the Euro area during the financial crisis, while Italy dominated during the sovereign crisis. The French corporates Groupama, Credit Industriel and Caisse d'Epargne were central in the inter-institutional contagion in both crises.

Recognition of protein allosteric states and residues: Machine learning approaches.

Allostery is a process by which proteins transmit the effect of perturbation at one site to a distal functional site upon certain perturbation. As an intrinsically global effect of protein dynamics, it is difficult to associate protein allostery with individual residues, hindering effective selection of key residues for mutagenesis studies. The machine learning models including decision tree (DT) and artificial neural network (ANN) models were applied to develop classification model for a cell signaling allosteric protein with two states showing extremely similar tertiary structures in both crystallographic structures and molecular dynamics simulations. Both DT and ANN models were developed with 75% and 80% of predicting accuracy, respectively. Good agreement between machine learning models and previous experimental as well as computational studies of the same protein validates this approach as an alternative way to analyze protein dynamics simulations and allostery. In addition, the difference of distributions of key features in two allosteric states also underlies the population shift hypothesis of dynamics-driven allostery model. © 2018 Wiley Periodicals, Inc.

Data Mining and Neural Networks Based Self-Adaptive Protection Strategies for Distribution Systems with DGs and FCLs

In light of the development of renewable energy and concerns over environmental protection, distributed generations (DGs) have become a trend in distribution systems. In addition, fault current limiters (FCLs) may be installed in such systems to prevent the short-circuit current from exceeding the capacity of the power apparatus. However, DGs and FCLs can lead to problems, the most critical of which is miscoordination in protection system. This paper proposes overcurrent protection strategies for distribution systems with DGs and FCLs. Through the proposed approach, relays with communication ability can determine their own operating states with the help of an operation setting decision tree and topology-adaptive neural network model based on data processed through continuous wavelet transform. The performance and effectiveness of the proposed protection strategies are verified by the simulation results obtained from various system topologies with or without DGs, FCLs, and load variations.

The application of data mining techniques to oral cancer prognosis.

This study adopted an integrated procedure that combines the clustering and classification features of data mining technology to determine the differences between the symptoms shown in past cases where patients died from or survived oral cancer. Two data mining tools, namely decision tree and artificial neural network, were used to analyze the historical cases of oral cancer, and their performance was compared with that of logistic regression, the popular statistical analysis tool. Both decision tree and artificial neural network models showed superiority to the traditional statistical model. However, as to clinician, the trees created by the decision tree models are relatively easier to interpret compared to that of the artificial neural network models. Cluster analysis also discovers that those stage 4 patients whose also possess the following four characteristics are having an extremely low survival rate: pN is N2b, level of RLNM is level I-III, AJCC-T is T4, and cells mutate situation (G) is moderate.

Design of the ZigBee Technology-based Wireless Sensor Network for Earth Temperature Monitoring

Aiming at the application of the experimental system of earth temperature monitoring, this study designed a ZigBee technology-based wireless sensor network (WSN) and designed the network topology according to actual demand. The system adopted a tree network model and used the CC2430 as the core chip of the wireless communication network of temperature measuring system. Moreover, DS18B20 temperature sensor was used to collect temperature data. According to the preset routing protocol, the temperature data obtained were uploaded to the coordinator node via the router node by multiple hops. The temperature data collected by coordinator node were transformed to the host after being processed. Applying the Cluster-Tree and AODVjr optimized routing algorithm, this system showed the advantages of low complexity, easy deployment, high efficiency, good real-time performance, low power consumption, and convenient for long-time work. In addition, it provided powerful support for the real-time monitoring of farmland experiment system.

Comparison of M5 Model Tree and Artificial Neural Network for Estimating Potential Evapotranspiration in Semi-arid Climates

Evaporation is a fundamental parameter in the hydrological cycle. This study examines the performance of M5 model tree and artificial neural network (ANN) models in estimating potential evapotranspiration calculated by Penman- Monteith and Hargreaves- Samani equations. Daily weather data from two meteorological stations in a semi-arid climate of Iran, namely Kerman and Zahedan, were collected during 1995-2004 and included the mean, maximum and minimum air temperatures, dewpoint, relative humidity, sunshine hours, and wind speed. Results for both stations showed that the performance of the M5 model tree was more accurate (R=0.982 and 0.98 for Penman-Monteith equation and R=0.983 and 0.98 for Hargreaves-Samani equation in Kerman and Zahedan, respectively) than the ANN model (R=0.975 and 0.978 for Penman-Monteith equation and R=0.967 and 0.974 for Hargreaves-Samani equation in Kerman and Zahedan, respectively), but the models’ differences were insignificant at a confidence level of 95%. It also performed better at the Zahedan station using the PenmanMonteith equation. The most significant variables affecting the potential evapotranspiration in the case of the Penman–Monteith equation were found to be mean air temperature, sunshine hours, wind speed, and relative humidity. Similarly, for the Hargreaves-Samani equation, the maximum and minimum temperatures, sunshine hours, and wind speed were determined to be the most significant variables. Further studies in other climates are recommended for further analysis.