Search

This paper presents a pioneer study of identifying the wind turbine blade condition based on its vibration pattern. The study used variational mode decomposition (VMD) for signal pre-processing. VMD is an adaptive signal decomposition technique, which can non-recursively decompose a multi-component signal into a number of quasi-orthogonal intrinsic mode functions. This new tool is based on a solid mathematical foundation and can obtain a well-defined time–frequency representation. The main advantage of VMD is, there is no residual noise in the modes and it can avoid redundant modes. Thus, VMD provides a noise-free fault signal for classification. Initially, the obtained vibration data are pre-processed using VMD and then the descriptive statistical features like mean, standard deviation, sample variance, and kurtosis were extracted. After the feature extraction, the dominant features were selected using the C4.5 decision tree algorithm. In this study, multilayer perceptron (MLP) was used for fault classification (blade bend, blade crack, erosion, hub-blade loose connection, and pitch angle twist). Then a comparative study was made with MLP without VMD and MLP with VMD to suggest a better model. From the results, the maximum classification accuracy of 87% was obtained for MLP with VMD when compared to MLP without VMD (76.83%).

In the past decades, the emerging concern about food safety has led to the increasing demand for monitoring food quality across the world. Aiming towards a novel solution for monitoring food, this study proposes a non-destructive method with self-powering capability for online food monitoring, which can be extendable to IoT applications. Furthermore, the study introduces a novel deep neural network model to predict different states of food quality based on the monitoring results. To monitor the variation in food quality, the paper proposes the detection of total volatile organic compounds (TVOCs) inside the food packages, which have been released during food deterioration. A low-power sensor mote comprised of a capacity humidity sensor and a metal-oxide (MOX) gas sensor was manufactured for this purpose. The self-powering capability of the mote is provided through an energy harvester module, which benefits from the far-field Radio Frequency Energy Harvesting (RFEH) technology. The operating frequency of the module was chosen at the 915-MHz ISM band. The analysis of the harvester performance showed that the harvester could generate 3.3-V dc with an RF input power of as low as -8 dBm, which was sufficient for the mote operation. To verify the proposed solutions, a demonstration to monitor the deterioration of packaged pork and fish was conducted in eight days under ambient and refrigerated storage conditions, using the self-developed RF-powered sensor mote. The raw variations in TVOCs were analyzed to evaluate the reliability of the proposed TVOC-based method. A one-dimensional (1-D) convolutional neural network (CNN) model was trained on the TVOCs dataset to predict different states of food quality. To investigate the applicability of the proposed 1-D CNN to multi-class determination of food quality, two other supervised machine learning algorithms using 2-D inputs, including Multilayer Perceptron (MLP), and Support Vector Machine (SVM), are studied. Their classification accuracies based on the confusion matrix are identified and compared.

Accurate mapping of crop distribution on Earth's surface aids in predicting grain production. Pattern classification along with remote sensing imagery can facilitate traditional manual field measurement techniques using machine learning. With the rapid increase in satellite sensor resolution, the object-based classification paradigm has increasingly been applied. However, scale parameter selection is always a difficult part of the object-based classification. Based on ensemble learning, this study proposes a classification method using the multiscale object-based weighted method which includes manual digitizing of crop distribution in the southern region of Jishan County, Shanxi Province, China, applying Gaofen-2 (GF-2) images. This method initially uses estimations of the scale parameter (ESP) tool to select “good” scales, defined here as “preferred” scales, after which feature subsets are screened by each preferred scale as the input of multiple classifiers and classifies. Finally, all classification results are then fused. Our research results indicate that: 1) Feature importance values are sorted differently at different preferred scales; 2) accuracy differences become clear when different preferred scales are combined with different classifiers, and determining the “best” single appropriate scale is generally difficult; 3) accuracy of the multiscale weighted classification method is higher compared to the single preferred scale approach. Furthermore, ensemble learning can be achieved using this method on multiple scales and on multiple classifiers. With this method, procedures that necessitate the selection of segmentation scales and the selection and optimization of classifiers can be skipped altogether.

One of the most persistent challenges concerning network security is to build a model capable of detecting intrusions in network systems. The issue has been extensively addressed in uncountable researches and using various techniques, of which a commonly used technique is that based on detecting intrusions in contrast to normal network traffic and the classification of network packets as either normal or abnormal. However, the problem of improving the accuracy and efficiency of classification models remains open and yet to be resolved. This study proposes a new binary classification model for intrusion detection, based on hybridization of Artificial Bee Colony algorithm (ABC) and Dragonfly algorithm (DA) for training an artificial neural network (ANN) in order to increase the classification accuracy rate for malicious and non-malicious traffic in networks. At first the model selects the suitable biases and weights utilizing a hybrid (ABC) and (DA). Next, the neural network is retrained using these ideal values in order for the intrusion detection model to be able to recognize new attacks. Ten other metaheuristic algorithms were adapted to train the neural network and their performances were compared with that of the proposed model. In addition, four types of intrusion detection evaluation datasets were applied to evaluate the proposed model in comparison to the others. The results of our experiments have demonstrated a significant improvement in inefficient network intrusion detection over other classification methods.

With coexisting-cooperative-cognitive robots gradually appearing in daily life, an instinct and efficient human-robot interaction (HRI) is becoming more and more challenging and necessary. Surface electromyography (sEMG) signals, as one of mainstream manners of the interactions, are employed to predict human intentions. In this paper, to provide natural assistance for standing up and sitting down, sEMG signals acquired from active muscles of one’s lower limb are utilized to predict continuous movements. A temporally smoothed multilayer perceptron (MLP) regression scheme is proposed for continuous knee/ankle angles estimation by multi-channel sEMG signals. After correlation analyses of sEMG signals and movements, a series of linear and nonlinear regression models are trained to decode human intentions from pre-processed sEMG. Furthermore, to remove out local fluctuations of direct mappings, temporally smoothed techniques are further implemented as post-processings. In the experiments of standing up and sitting down, extensive results of ten healthy subjects show that a three-layer MLP with the Savitzky-Golay filter achieves the best performance on the mean squared error (MSE, testing: 59.58) and the $R^{2}$ score ( $R^{2}$ , testing: 0.948). The proposed regression scheme is compared with other methods and is also verified by measurements of a high-precision visual motion capture system.

Coronavirus is a pandemic that has become a concern for the whole world. This disease has stepped out to its greatest extent and is expanding day by day. Coronavirus, termed as a worldwide disease, has caused more than 8 lakh deaths worldwide. The foremost cause of the spread of coronavirus is SARS-CoV and SARS-CoV-2, which are part of the coronavirus family. Thus, predicting the patients suffering from such pandemic diseases would help to formulate the difference in inaccurate and infeasible time duration. This paper mainly focuses on the prediction of SARS-CoV and SARS-CoV-2 using the B-cells dataset. The paper also proposes different ensemble learning strategies that came out to be beneficial while making predictions. The predictions are made using various machine learning models. The numerous machine learning models, such as SVM, Naïve Bayes, K-nearest neighbors, AdaBoost, Gradient boosting, XGBoost, Random forest, ensembles, and neural networks are used in predicting and analyzing the dataset. The most accurate result was obtained using the proposed algorithm with 0.919 AUC score and 87.248% validation accuracy for predicting SARS-CoV and 0.923 AUC and 87.7934% validation accuracy for predicting SARS-CoV-2 virus.

The aim of this paper is twofold. First, black hole algorithm (BHA) is proposed as a new training algorithm for feedforward neural networks (FNNs), since most traditional and metaheuristic algorithms for training FNNs suffer from the problem of slow coverage and getting stuck at local optima. BHA provides a reliable alternative to address these drawbacks. Second, complementary learning components and Levy flight random walk are introduced into BHA to result in a novel optimization algorithm (BHACRW) for the purpose of improving the FNNs’ accuracy by finding optimal weights and biases. Four benchmark functions are first used to evaluate BHACRW’s performance in numerical optimization problems. Later, the classification performance of the suggested models, using BHA and BHACRW for training FNN, is evaluated against seven various benchmark datasets: iris, wine, blood, liver disorders, seeds, Statlog (Heart), balance scale. Experimental result demonstrates that the BHACRW performs better in terms of mean square error (MSE) and accuracy of training FNN, compared to standard BHA and eight well-known metaheuristic algorithms: whale optimization algorithm (WOA), biogeography-based optimizer (BBO), gravitational search algorithm (GSA), genetic algorithm (GA), cuckoo search (CS), multiverse optimizer (MVO), symbiotic organisms search (SOS), and particle swarm optimization (PSO). Moreover, we examined the classification performance of the suggested approach on the angiotensin-converting enzyme 2 (ACE2) gene expression as a coronavirus receptor, which has been overexpressed in human rhinovirus-infected nasal tissue. Results demonstrate that BHACRW-FNN achieves the highest accuracy on the dataset compared to other classifiers.

Data-driven methods open up unprecedented possibilities for maritime surveillance using Automatic Identification System (AIS) data. In this work, we explore deep learning strategies using historical AIS observations to address the problem of predicting future vessel trajectories with a prediction horizon of several hours. We propose novel sequence-to-sequence vessel trajectory prediction models based on encoder-decoder recurrent neural networks (RNNs) that are trained on historical trajectory data to predict future trajectory samples given previous observations. The proposed architecture combines Long Short-Term Memory (LSTM) RNNs for sequence modeling to encode the observed data and generate future predictions with different intermediate aggregation layers to capture space-time dependencies in sequential data. Experimental results on vessel trajectories from an AIS dataset made freely available by the Danish Maritime Authority show the effectiveness of deep-learning methods for trajectory prediction based on sequence-to-sequence neural networks, which achieve better performance than baseline approaches based on linear regression or on the Multi-Layer Perceptron (MLP) architecture. The comparative evaluation of results shows: i) the superiority of attention pooling over static pooling for the specific application, and ii) the remarkable performance improvement that can be obtained with labeled trajectories, i.e., when predictions are conditioned on a low-level context representation encoded from the sequence of past observations, as well as on additional inputs (e.g., port of departure or arrival) about the vessel's high-level intention, which may be available from AIS., Accepted for publications in IEEE Transactions on Aerospace and Electronic Systems, 17 pages, 9 figures

In recent years, the development and influence of wind power in the power system have witnessed, which has led to a significant increase in the production and use of wind energy worldwide. Considering the variability of wind velocity, planning, and managing wind intermittency are important parts of wind energy development, so predicting wind speeds for high-efficiency energy production is one of the most important power system planning issues. Nowadays, machine learning methods are widely used to model complex and nonlinear systems such as wind speed or solar radiation. In this chapter, wind speed prediction models using machine learning applications are presented to solve power system planning problems. This study utilized two machine learning applications called multilayer perceptron (MLP) and group method of data handling (GMDH) to predict wind speed. To evaluate the proposed models, the authors will predict the wind speed for 15 months as a short-term wind speed prediction. Wind speed prediction in the 15 months horizon is done hourly for each day. The presented results illustrate the proposed models’ capability and effectiveness for predicting short-term wind speeds based on historical wind speed data and the good correlation between the predicted and actual values of data. Wind speed forecasting and wind resource assessment can show the right investment direction to decision-makers and investors, thereby developing the wind energy industry and creating a sustainable power system.

The accurate prediction of the solar Diffuse Fraction (DF), sometimes called the Diffuse Ratio, is an important topic for solar energy research. In the present study, the current state of Diffuse Irradiance research is discussed and then three robust, Machine Learning (ML) models, are examined using a large dataset (almost 8 years) of hourly readings from Almeria, Spain. The ML models used herein, are a hybrid Adaptive Network-based Fuzzy Inference System (ANFIS), a single Multi-Layer Perceptron (MLP) and a hybrid Multi-Layer Perceptron-Grey Wolf Optimizer (MLP-GWO). These models were evaluated for their predictive precision, using various Solar and Diffuse Fraction (DF) irradiance data, from Spain. The results were then evaluated using two frequently used evaluation criteria, the Mean Absolute Error (MAE) and the Root Mean Square Error (RMSE). The results showed that the MLP-GWO model, followed by the ANFIS model, provided a higher performance, in both the training and the testing procedures., Comment: 13 pages, 6 figures

Both statistical and neural methods have been proposed in the literature to predict healthcare expenditures. However, less attention has been given to comparing predictions from both these methods as well as ensemble approaches in the healthcare domain. The primary objective of this paper was to evaluate different statistical, neural, and ensemble techniques in their ability to predict patients' weekly average expenditures on certain pain medications. Two statistical models, persistence (baseline) and autoregressive integrated moving average (ARIMA), a multilayer perceptron (MLP) model, a long short-term memory (LSTM) model, and an ensemble model combining predictions of the ARIMA, MLP, and LSTM models were calibrated to predict the expenditures on two different pain medications. In the MLP and LSTM models, we compared the influence of shuffling of training data and dropout of certain nodes in MLPs and nodes and recurrent connections in LSTMs in layers during training. Results revealed that the ensemble model outperformed the persistence, ARIMA, MLP, and LSTM models across both pain medications. In general, not shuffling the training data and adding the dropout helped the MLP models and shuffling the training data and not adding the dropout helped the LSTM models across both medications. We highlight the implications of using statistical, neural, and ensemble methods for time-series forecasting of outcomes in the healthcare domain.

We present a machine learning based COVID-19 cough classifier which can discriminate COVID-19 positive coughs from both COVID-19 negative and healthy coughs recorded on a smartphone. This type of screening is non-contact, easy to apply, and can reduce the workload in testing centres as well as limit transmission by recommending early self-isolation to those who have a cough suggestive of COVID-19. The datasets used in this study include subjects from all six continents and contain both forced and natural coughs, indicating that the approach is widely applicable. The publicly available Coswara dataset contains 92 COVID-19 positive and 1079 healthy subjects, while the second smaller dataset was collected mostly in South Africa and contains 18 COVID-19 positive and 26 COVID-19 negative subjects who have undergone a SARS-CoV laboratory test. Both datasets indicate that COVID-19 positive coughs are 15\%-20\% shorter than non-COVID coughs. Dataset skew was addressed by applying the synthetic minority oversampling technique (SMOTE). A leave-$p$-out cross-validation scheme was used to train and evaluate seven machine learning classifiers: LR, KNN, SVM, MLP, CNN, LSTM and Resnet50. Our results show that although all classifiers were able to identify COVID-19 coughs, the best performance was exhibited by the Resnet50 classifier, which was best able to discriminate between the COVID-19 positive and the healthy coughs with an area under the ROC curve (AUC) of 0.98. An LSTM classifier was best able to discriminate between the COVID-19 positive and COVID-19 negative coughs, with an AUC of 0.94 after selecting the best 13 features from a sequential forward selection (SFS). Since this type of cough audio classification is cost-effective and easy to deploy, it is potentially a useful and viable means of non-contact COVID-19 screening., Comment: This paper has been accepted in "Computers in Medicine and Biology" and currently under production

Phishing is a type of Internet fraud that aims to acquire the credential of users via scamming websites. In this paper, a novel approach is utilized that uses a Neural Network with a multilayer perceptron to detect the scam URL. The proposed system improves the accuracy of the scam detection system as it achieves a high accuracy percentage of 98.5%.

The increasing liberalization of European electricity markets, the growing proportion of intermittent renewable energy being fed into the energy grids, and also new challenges in the patterns of energy consumption (such as electric mobility) require flexible and intelligent power grids capable of providing efficient, reliable, economical, and sustainable energy production and distribution. From the supplier side, particularly, the integration of renewable energy sources (e.g., wind and solar) into the grid imposes an engineering and economic challenge because of the limited ability to control and dispatch these energy sources due to their intermittent characteristics. Time-series prediction of wind speed for wind power production is a particularly important and challenging task, wherein prediction intervals (PIs) are preferable results of the prediction, rather than point estimates, because they provide information on the confidence in the prediction. In this paper, two different machine learning approaches to assess PIs of time-series predictions are considered and compared: 1) multilayer perceptron neural networks trained with a multiobjective genetic algorithm and 2) extreme learning machines combined with the nearest neighbors approach. The proposed approaches are applied for short-term wind speed prediction from a real data set of hourly wind speed measurements for the region of Regina in Saskatchewan, Canada. Both approaches demonstrate good prediction precision and provide complementary advantages with respect to different evaluation criteria.

WOS:000301404600026 This study investigates the potential of Time Lag Recurrent Neural Networks (TLRN) for modeling the daily inflow into Eleviyan reservoir, Iran. TLRN are extended with short term memory structures that have local recurrent connections, thus making them an appropriate model for processing temporal (time-varying) information. For this study, the daily inflow into Eleviyan reservoir between years 2004-2007 was considered. To compare the performance of TLRN, a back propagation neural network was used. The TLRN model with gamma memory structure, eight input layer nodes, two hidden layer and one output layer (8-2-1) was found performing best out of three different models used in forecasting daily inflow. A comparison of results with back propagation neural network suggest that neither TLRN nor back propagation approaches were good in forecasting high inflow but, both approaches perform well when used to forecast low inflow values. However, statistical test suggests that both TLRN and back propagation neural network models were able to reproduce similar basic statistics as that of the actual data. (C) 2011 Elsevier Inc. All rights reserved.

PubMedID: 18814496 Epilepsy is a disorder of cortical excitability and still an important medical problem. The correct diagnosis of a patient's epilepsy syndrome clarifies the choice of drug treatment and also allows an accurate assessment of prognosis in many cases. The aim of this study is to evaluate epileptic patients and classify epilepsy groups such as partial and primary generalized epilepsy by using Radial Basis Function Neural Network (RBFNN) and Multilayer Perceptron Neural Network (MLPNNs). Four hundred eighteen patients with epilepsy diagnoses according to International League against Epilepsy (ILAE 1981) were included in this study. The correct classification of this data was performed by two expert neurologists before they were executed by neural networks. The neural networks were trained by the parameters obtained from the EEG signals and clinic properties of the patients. Experimental results show that the predictions of both neural network models are very satisfying for learning data sets. According to test results, RBFNN (total classification accuracy=95.2%) has classified more successfully when compared with MLPNN (total classification accuracy=89.2%). These results indicate that RBFNN model may be used in clinical studies as a decision support tool to confirm the classification of epilepsy groups after the model is developed. © 2008 Springer Science+Business Media, LLC.

International audience; On one hand, sparse coding, which is widely used in signal processing, consists of representing signals as linear combinations of few elementary patterns selected from a dedicated dictionary. The output is a sparse vector containing few coding coefficients and is called sparse code. On the other hand, Multilayer Perceptron (MLP) is a neural network classification method that learns non linear borders between classes using labeled data examples. The MLP input data are vectors, usually normalized and preprocessed to minimize the inter-class correlation. This article acts as a link between sparse coding and MLP by converting sparse code into convenient vectors for MLP input. This original association assures in this way the classification of any sparse signals. Experimental results obtained by the whole process on trajectories data and comparisons to other methods show that this approach is efficient for signals classification.

Complex network systems are pervasive in life sciences at all levels, from molecules and genes to organisms and ecosystems. All these systems are characterized by being constituted of numerous components or nodes (molecules, genes, cells, tissues, organisms), which are interconnected by many links in an intricate tangle, just as biological neural networks consist of many interacting neurons (Fig. 1). Apart from its structural complexity, complex networks are inherently difficult to understand because interactions are non-linear, distributed non-randomly, and are adaptive, that is, changing continuously in response to the state of the system itself (Strogatz, 2001; Pascual & Dunne, 2006). Understanding the functioning of these systems consisting of a large number of strongly interacting units represents therefore a major endeavour for biologists and ecologists. As complex networks, ecosystems are non-linear systems constituted by countless interacting pieces, both biotic and abiotic, constituting the entangled web of life. In a world threatened by global environmental problems such as biodiversity loss, climate change, fishing overexploitation or pollution, ecologists are challenged by the need to understand and predict the dynamics of ecosystems as never before. Along with the complexity of ecological systems, ecologists are also faced with a huge amount of information that recent advances in data collection technology such as remote sensing have produced. To cope with the ecosystem complexity and large data sets currently available, ecologists nowadays have the opportunity to use machine-like learning techniques such as the artificial neural networks (ANNs). As their name implies, ANNs are biologically inspired and were initially intended to mimic the neural activity in the human or animal brains (Garson, 1991; Goh, 1995; Stern, 1996). ANNs models are based on the same learning processes as the animal brain, which gathers information from the environment (input data) and gives an answer (output data) after using learned training algorithms. However, given that the architecture and dynamics of the animal brain is exceedingly complex, even the most elaborated ANN models are mere caricatures of the biological brain. Although the original works on ANNs date back to the forties (McCulloch & Pitts, 1943; Pitts & McCulloch, 1947), they not became really popular until the eighties after the work of the physicist John Hopfield. Hopfield (1982) introduced an oversimplified neural network, comprising a set of fully connected binary units, as a metaphor of neural computation. The most remarkable feature of this model was that it could learn by association and was quiet insensitive to noise. This capacity to recognize previously learned patterns, which was thought to be an exclusive property of brains, is precisely what the Hopfield model does (Sole & Goodwin, 2000).

This study developed 12 prediction models using two types of data matrix (daily means and a selection of the mean for the first 6 h of the day). The Persistence parametric prediction technique was applied separately to these matrices, as well as semiparametric Ridge Regression and three non-parametric or artificial intelligence techniques: Support Vector Machine, Multilayer Perceptron and ELMAN networks. The target was the prediction of maximum tropospheric ozone concentrations for the next day in the Mexicali-Calexico border area. The main ozone precursors and meteorological parameters were used for the different models. The proposals were evaluated using specific performance measurements for the air quality models established in the Model Validation Kit and recommended by the US Environmental Protection Agency. Results with similar margins of error were obtained in various models developed in this study, and some of them have provided smaller margins of error than similar prediction models existing in the literature developed in other regions. For this reason, we consider it feasible to apply the prediction models developed and they could be useful for supporting decisions in the matter of ozone pollution in the region under study, as well as for use in daily forecasting in this area. © 2007 Elsevier Ltd. All rights reserved.

The daily average PM 2.5 concentration forecast is a leading component nowadays in air quality research, which is necessary to perform in order to assess the impact of air on the health and welfare of every living being. The present work is aimed at analyzing and benchmarking a neural-network approach to the prediction of average PM 2.5 concentrations. The model thus obtained will be indispensable, as a control tool, for the purpose of preventing dangerous situations that may arise. To this end we have obtained data and measurements based on samples taken during the early hours of the day. Results from three different topologies of neural networks were compared so as to identify their potential uses, or rather, their strengths and weaknesses: Multilayer Perceptron (MLP), Radial Basis Function (RBF) and Square Multilayer Perceptron (SMLP). Moreover, two classical models were built (a persistence model and a linear regression), so as to compare their results with the ones provided by the neural network models. The results clearly demonstrated that the neural approach not only outperformed the classical models but also showed fairly similar values among different topologies. Moreover, a differential behavior in terms of stability and length of the training phase emerged during testing as well. The RBF shows up to be the network with the shortest training times, combined with a greater stability during the prediction stage, thus characterizing this topology as an ideal solution for its use in environmental applications instead of the widely used and less effective MLP. © 2004 Elsevier Ltd. All rights reserved.

In pattern recognition, it is well known that the classifier performance depends on the classification rule and the complexities presented in the data sets (such as class overlapping, class imbalance, outliers, high-dimensional data sets among others). In this way, the issue of class imbalance is exhibited when one class is less represented with respect to the other classes. If the classifier is trained with imbalanced data sets, the natural tendency is to recognize the samples included in the majority class, ignoring the minority classes. This situation is not desirable because in real problems it is necessary to recognize the minority class more without sacrificing the precision of the majority class. In this work we analyze the behaviour of four classifiers taking into a count a relative balance among the accuracy classes.