Multilayer Perceptron Neural Network Methods Research Articles

A hybrid data-driven framework for loss prediction of MCA airfoils

The Multi Circular Arc airfoils family has gained significant popularity in axial compressor design due to its capability to achieve higher pressure ratios with lower losses compared to conventional NACA airfoils. However, modeling their aerodynamic performance for transonic axial compressors remains a challenging task. This research addresses this issue by developing a data driven model for Multi Circular Arc airfoils loss correlation utilizing an automated blade-to-blade flow solver, StarCCM, under specific flow boundary conditions. Since the data built in this research includes high dimensionality, it is subjected to dimensionality reduction using Principal Component Analysis while retaining as much of its important information as possible. What sets our research apart is the comparative analysis of hybrid machine learning approaches. We specifically combine classification techniques based on shock wave formation on the airfoil’s suction side with Multi-Layer Perceptron neural network methods. This unique combination allows for a comprehensive prediction of MCA loss as the main objective of the research. The results demonstrate that this surrogate-based approach provides accurate loss model with 97% of precision error for the classification which results in 0.99 and 0.94 of R2 for loss model related to the transonic and subsonic airfoils loss prediction.

A multi-energy load prediction of a building using the multi-layer perceptron neural network method with different optimization algorithms

Since cooling and heating loads are recognized as key characteristics for evaluating the energy efficiency of buildings, it appears indisputable that they must be predicted and analyzed for residential structures. Accordingly, the multi-layer perceptron neural network is applied for predicting the heating and cooling loads using the experimental dataset. The used dataset is obtained by monitoring the impact of the building's dimensions on energy consumption. To optimize the training process of the multi-layer perceptron neural network, several optimizers are employed. Besides, different statistical performance indicators are considered to see which selected optimizer outperforms the rest in terms of accuracy and authenticity. The obtained results emphasize the remarkable performance of adaptive chaotic grey wolf optimization, which can be used to train the multi-layer perceptron neural network and forecast the buildings’ energy consumption with the highest accuracy. According to the obtained results, the hybrid multi-layer perceptron neural network- adaptive chaotic grey wolf optimization method demonstrates the best performance. The optimum number of neurons in the hidden layer is obtained to be 15. Also, based on the statistical performance indicators, the selected method reveals an R2 of 0.9123 and 0.9419 for cooling and heating loads, respectively.

Computational prediction of the drilling rate of penetration (ROP): A comparison of various machine learning approaches and traditional models

Rate of penetration (ROP) prediction, can assist precise planning of drilling operations and can reduce drilling costs. However, easy estimation of this key factor by traditional or experimental models is very difficult. This requires comparing available models for achieving the best prediction approach. In this study, four machine learning (ML) methods and two traditional ROP models were utilized to predict ROP. ML techniques include multilayer perceptron neural network (MLPNN), radial basis function neural network (RBFNN), adaptive neuro-fuzzy inference system (ANFIS), and support vector regression (SVR). MLPNN, ANFIS, and RBFNN methods were trained with four meta-heuristic algorithms including particle swarm optimization (PSO), ant colony optimization (ACO), differential evolution (DE), and genetic algorithm (GA). The backpropagation (BP) algorithm was also incorporated to train the ANFIS and MLPNN methods as a conventional method. For comparison purposes, the traditional ROP models of Bourgoyne and Young (BYM) and Bingham were also implemented in combination with four meta-heuristic algorithms. The required data were collected from the mud logging unit (MLU) and the final report of a drilled well located in southwestern of Iran. In the MLU, the information of different sensors is firstly collected through the data communications protocols and sent to a master unit to be processed into relevant dimensions (i.e. create operational variables). In order to accurately record information, the sensors are calibrated at regular intervals. After removing the outliers, the overall noise of data was reduced by Savitzky-Golay (SG) smoothing filter. Then, in order to simulate a drilling process in a realistic manner and also to evaluate the performance of the models in approximating the penetration rate at greater depth of hole, the available data were divided into 6 sections based on depth. After that, sections 2 Data collection and preprocessing , 3 Methodology , 4 Model development and ROP estimation , 5 Discussion and results , 6 Conclusion were separately used as a test dataset and all previous sections were considered as a training dataset. Results indicated that PSO-MLPNN achieved the highest performance in comparing with other developed models. The concluding remark is that ML models are more efficient and reliable than traditional models. In addition, combining ML models with meta-heuristic algorithms can achieve better results than conventional algorithms such as BP. The results of this study can be used as a practical guide for the management and planning of future well drilling. • Applying Savitzky-Golay (SG) filter to remove noise from data. • Data were divided into 6 sections. Section 2 to 6 were separately noted as test and the previous one(s) as training dataset. • Comparison of four machine learning (ML) methods and two traditional ROP models. • Coupling models with meta-heuristic algorithms and conventional algorithms. • PSO-MLPNNmodel showed the highest accuracy compared to others.

Improving Dropout Forecasting during the COVID-19 Pandemic through Feature Selection and Multilayer Perceptron Neural Network

Nowadays, online education in universities is mature from the situation of COVID-19 spread. It has greatly changed the learning environment in the classroom and has also resulted in students’ dropout. The purpose of this study is to examine the factors that influence students’ dropout in the department of information technology of Suan Sunandha Rajabhat University. Students enrolled in the subjects in the COVID-19 pandemic situation must study online at their homes and must take online courses. This study investigated 1,650 student records, 19,450 enrollments, 16,200 grades, and 11,780 social media accounts that had access to all online courses. It was examined how to improve the model's performance by combining feature selection with a multilayer perceptron neural network method. The model was compared to student dropout predictions generated by Logistic regression, Decision Tree, Random Forest, Naive Bayes, Support Vector Machine, and Multilayer Perceptron Neural Network, with feature selection (1). The 10-Folds Cross Validation method was used to determine the efficiency of the Gain Ratio, Chi-Square, and Correlation-based Feature Selection models to compare accuracy, precision, sensitivity, F1 score, and classification error rate (e). After adjusting the modeling parameters, the multilayer perceptron neural network method combined with CFS characterization achieved an accuracy of 96.98%. The study’s findings indicate that the feature selection technique can be used to improve the neural network model's efficiency in predicting student dropout during a COVID-19 pandemic. Furthermore, the simulation can improve student dropout forecasting during spread out that persists.

Novel neural network optimization approach for modeling scattering and noise parameters of microwave transistor

AbstractThis study performed modeling of the scattering (S) and noise (N) parameters of the ATF53189 using the General Regression Neural Network (GRNN) and Multi Layer Perceptron Neural Network (MLPNN) methods based on Artificial Neural Network (ANN). For modeling the linear behavior of the transistor, the optimum design parameters of the GRNN and the MLPNN methods were determined using four different optimization algorithms. These are whale optimization algorithm (WOA), artificial bee colony (ABC), particle swarm optimization (PSO) and ant lion optimizer (ALO) algorithms. With the help of these algorithms, the sigma parameter of the GRNN and the number of hidden layers, numbers of neurons in the hidden layers and the activation functions of the hidden layers of the MLPNN were optimized. This way, the best models required for prediction of the S and N parameters of the ATF53189 were obtained. Different models that provided each of the angles and magnitudes of the S11, S21, S12, S22 parameters and the Fmin, Γmin, Γopt magnitude, Γopt angle and Rn noise parameters as the output were created. The experimental results showed that the GRNN method should be used in linear behavior modeling of S parameters of the ATF53189 and the MLPNN method should be used in linear behavior modeling of N parameters of the ATF53189. It is understood that the best algorithm for optimizing the design parameters of the GRNN and the MLPNN methods is the PSO. As a result, the modeling of the S and N parameters of the ATF53189 transistor was successfully carried out with the methods used in this study.

Framework for agricultural e-trading platform adoption using neural networks

The digital e-trading platform is an ambitious and priority intervention in the agricultural sector. The research study is among the very few that identifies and prioritizes the predictors of wholesale electronic trading platform adoption in agricultural marketing. The research has been conducted in two phases—identification of predictors based on a scholarly articles’ review and generating the predictive function between the identified predictors and adoption using the multilayer perceptron neural networks method. A case study of the electronic national agriculture market platform is undertaken to understand the predictors in the context of a real project. The study has revealed significant predictors of priority (high to low) as fast transaction cycles, higher prices, easier to use, infrastructure availability, trust, social influence, low transaction costs and customer care. Post-adoption, the user expects to get benefits like transparency, quick trade settlement, reduced transaction cost, expanding the market reach, and increase in the product price realized by the farmer. The improvement in the critical predictors of the electronic trading platform may help the de-facto national initiative to succeed. In new markets, the prediction framework using the neural network may be used to identify users with the propensity to adopt e-commerce or digital initiatives in the agricultural sector and proactively approach them to get a high number of transactions—an essential ingredient of success. The digital e-trading initiative is a growth catalyst of Agriculture 4.0 and may transform the agricultural supply chain for the better.

A study on the behavior of CO2 corrosion on pipeline using computational fluid dynamics, experimental and artificial neural network approach

Corrosion of the piping system is a genuine problem in the oil and gas industry. Most oil and gas industries used a carbon steel pipeline for the transportation of crude oil, which is affected by CO2 corrosion. Now a day, the computational approach and artificial neural network approach will be used to study the corrosion rate. Therefore, in this work, Computational Fluid Dynamics (CFD) and Artificial Neural Network (ANN) studies on piping systems were made to determine the corrosion rate induced by CO2 saturated aqueous solutions on carbon steel pipeline. In CFD study, corrosion rates were computed by modeling the electrochemical processes occurring at the metal substrate from cathodic reductions of the carbonic acid and hydrogen ions, and the anodic oxidation of the metal component. Also, an artificial neural network study was made using a multilayer perceptron neural network method; and, computational fluid dynamics and artificial neural network simulations were validated with in-house built experiment set-up. The experimental study had been carried out for more than 200-h to find the corrosion rate on the pipeline, and satisfactory trends were observed between computational fluid dynamics, artificial neural network, and experimental values. In the end, corroded pipes were observed under a scanning electron microscope and x-ray spectroscopy, and the corroded zones were viewed as against the non-corroded pipe.

Application of Group Method of Data Handling Type Neural Network for Significant Wave Height Prediction

The estimation of wave parameters is of great importance in coastal activities such as design studies for harbor, inshore and offshore structures, coastal erosion, sediment transport, and wave energy estimation. For this purpose, several models and approaches have been proposed to predict wave parameters, such as empirical, numerical-based approaches, and soft computing. In this study, the group method of data handling type neural network (GMDH-NN) was presented for significant wave height prediction in an attempt to suggest a new model with superior explanatory power and stability. The GMDH-NN results were compared with the field data and with a multilayer perceptron neural networks (MLPNN) model. The results indicate that the prediction accuracy and avoidance of over-fitting of the GMDH-NN method were superior to those of the MLPNN method. The percentage improvement in the root mean square error and the mean absolute percentage error of the GMDH-NN model over the MLPNN model were 72.92% and 81.02%, respectively. Also, according to the indices, the GMDH-NN model performs the best for predicting the Hs of all of the wave height ranges. That is, the GMDH-NN model is capable of predicting wave heights for different ranges. The results of the analysis suggest that the GMDH-NN-based modeling is effective in predicting significant wave height.

Improving the Diagnosis of Liver Disease Using Multilayer Perceptron Neural Network and Boosted Decision Trees

Early detection of liver disease is never easy, though it is one of the most important diseases on earth. This study, thus, attempts to achieve efficient early detection through a Multilayer Perceptron Neural Network (MLPNN) algorithm based on various decision tree algorithms such as See5 (C5.0), Chi square Automatic interaction detector (CHAID) and classification and regression tree (CART) with boosting technique. Five hundred and eighty-three records related to the Indian Liver Patient Dataset (ILPD) were collected from University of California, Irvine (UCI) repository dataset for the verification of the proposed work. The ILPD dataset is divided into 70% for the training stage and 30% for the testing stage. Several evaluation metrics, such as specificity, sensitivity, precision, false positive rate (FPR), false negative rate (FNR), F1, and accuracy, are applied in this study. These metrics are carried out in two phrases. In the first experiment, we observed that B-C5.0 method presents better performance than B-CHAID and B-CART methods. In the second experiment, a hybridization of B-C5.0 and MLPNN methods, namely MLPNNB-C5.0, indicates the highest rates of detection of liver disease compared to other algorithms. Results indicate that MLPNNB-CHAID method has the most innovative accuracy with a value of 14.57%. The proposed method is able to diagnose and classify the liver disease efficiently. We would argue that the proposed system can be useful as a medical data mining approach in order to provide an early diagnosis of liver disease.

Respiratory signal prediction based on adaptive boosting and multi-layer perceptron neural network

To improve the prediction accuracy of respiratory signals using adaptive boosting and multi-layer perceptron neural network (ADMLP-NN) for gated treatment of moving target in radiation therapy. The respiratory signals acquired using a real-time position management (RPM) device from 138 previous 4DCT scans were retrospectively used in this study. The ADMLP-NN was composed of several artificial neural networks (ANNs) which were used as weaker predictors to compose a stronger predictor. The respiratory signal was initially smoothed using a Savitzky–Golay finite impulse response smoothing filter (S–G filter). Then, several similar multi-layer perceptron neural networks (MLP-NNs) were configured to estimate future respiratory signal position from its previous positions. Finally, an adaptive boosting (Adaboost) decision algorithm was used to set weights for each MLP-NN based on the sample prediction error of each MLP-NN. Two prediction methods, MLP-NN and ADMLP-NN (MLP-NN plus adaptive boosting), were evaluated by calculating correlation coefficient and root-mean-square-error between true and predicted signals. For predicting 500 ms ahead of prediction, average correlation coefficients were improved from 0.83 (MLP-NN method) to 0.89 (ADMLP-NN method). The average of root-mean-square-error (relative unit) for 500 ms ahead of prediction using ADMLP-NN were reduced by 27.9%, compared to those using MLP-NN. The preliminary results demonstrate that the ADMLP-NN respiratory prediction method is more accurate than the MLP-NN method and can improve the respiration prediction accuracy.

Enhanced RBF neural network model for time series prediction of solar cells panel depending on climate conditions (temperature and irradiance)

A radial basis function neural network is an effective technique for function approximation and prediction. It has been used in many applications in the real world; one of them is the time series prediction which is a relatively complex problem. In this paper, we propose an enhanced radial basis function neural network (RBFNN) model that depends on the standard RBF built-in MATLAB (newrb). The enhancement on newrb depends on the use of intelligent algorithms like K-means clustering, K-nearest neighbor, and singular value decomposition, to optimize the centers c, radii r, and weights w of the RBFNN. These algorithms replace the mathematical calculation used to find these parameters in newrb. The proposed enhanced model is applied to predict the solar cells energy production in Palestine using already installed solar panels in Jericho city. Solar irradiance and daily temperature are used as an input training data set for the proposed model, with the real output power of (2015) as the training supervisor. The model is applied to predict the output power within 1 month and for 1 year. Finally, a power output equation was optimized to calculate the solar energy depending on the irradiance and temperature with an acceptable accuracy. The experimental results show that the enhanced model performs more precisely than the traditional RBFNN and multilayer perceptron neural network methods, with low mean square error of relatively few neurons on the hidden layer.

A wavelet based approach for combining the outputs of different rainfall–runoff models

The rainfall–runoff modelling being a stochastic process in nature is dependent on various climatological variables and catchment characteristics and therefore numerous hydrological models have been developed to simulate this complex process. One approach to modelling this complex non-linear rainfall–runoff process is to combine the outputs of various models to get more accurate and reliable results. This multi-model combination approach relies on the fact that various models capture different features of the data, and hence combination of these features would yield better result. This study for the first time presented a novel wavelet based combination approach for estimating combined runoff The simulated daily output (Runoff) of five selected conventional rainfall–runoff models from seven different catchments located in different parts of the world was used in current study for estimating combined runoff for each time period. Five selected rainfall–runoff models used in this study included four data driven models, namely, the simple linear model, the linear perturbation model, the linearly varying variable gain factor model, the constrained linear systems with a single threshold and one conceptual model, namely, the soil moisture accounting and routing model. The multilayer perceptron neural network method was used to develop combined wavelet coupled models to evaluate the effect of wavelet transformation (WT). The performance of the developed wavelet coupled combination models was compared with their counterpart simple combination models developed without WT. It was concluded that the presented wavelet coupled combination approach outperformed the existing approaches of combining different models without applying input WT. The study also recommended that different models in a combination approach should be selected on the basis of their individual performance.

Bandwidth extension of narrowband speech in log spectra domain using neural network

In recent years, there have been significant advances in communication technology, but speech signals still suffer from low perceived quality caused by bandwidth limitations of telephone networks. The bandwidth extension (BWE) approach adds high-frequency components of the speech signal to band-limited telephone speech and increases speech perception significantly. In this work, we develop a new method for representation of vocal tract filter coefficients using log of filter bank energy (LFBE) parameters as an alternative for mel-frequency cepstral coefficients (MFCCs). This approach is based on a strong correlation between the spectral components of low- and high-band spectrums. Furthermore, the performances of Gaussian mixture model and multilayer perceptron neural network methods for estimation of the high-frequency envelope are evaluated. Objective evaluations of the obtained results indicate that the LFBE feature vectors have better performance than the MFCCs. In addition, findings of the objective evaluations showed that using a neural network, which is not common in BWE, achieves a better performance as compared to the Gaussian mixture model.

Polynomials, radial basis functions and multilayer perceptron neural network methods in local geoid determination with GPS/levelling

The means for determining reference surfaces when using satellite-based positioning techniques differs from the way they are determined using classical terrestrial positioning techniques. The height of any point on earth, as determined by a global positioning system (GPS), is based on the World Geodetic System of 1984 datum, but when using classical terrestrial measurements this height is based on the geoid. However, GPS-derived ellipsoidal heights must initially be transformed into orthometric heights for practical applications. We use geoid models developed with geoid heights for the height transformation, because orthometric height determination methods using classical techniques require time and manpower. In this study, the performances of polynomials, radial basis functions (RBFs) and multilayer perceptron (MLP) neural network algorithms used in local geoid surface modelling were evaluated in the study area. Upon analysis of statistical results, the artificial neural network method was observed to give better results than the other two methods.

A new power system transient stability assessment method based on Type-2 fuzzy neural network estimation

Transient stability assessment (TSA) of large power systems by the conventional method is a time consuming task. For each disturbance many nonlinear equations should be solved that makes the problem too complex and will lead to delayed decisions in providing the necessary control signals for controlling the system. Nowadays new methods which are devise artificial intelligence techniques are frequently used for TSA problem instead of traditional methods. Unfortunately these methods are suffering from uncertainty in input measurements. Therefore, there is a necessity to develop a reliable and fast online TSA to analyze the stability status of power systems when exposed to credible disturbances. We propose a direct method based on Type-2 fuzzy neural network for TSA problem. The Type-2 fuzzy logic can properly handle the uncertainty which is exist in the measurement of power system parameters. On the other hand a multilayer perceptron (MLP) neural network (NN) has expert knowledge and learning capability. The proposed hybrid method combines both of these capabilities to achieve an accurate estimation of critical clearing time (CCT). The CCT is an index of TSA in power systems. The Type-2 fuzzy NN is trained by fast resilient back-propagation algorithm. Also, in order to the proposed approach become scalable in a large power system, a NN based sensitivity analysis method is employed to select more effective input data. Moreover, In order to verify the performance of the proposed Type-2 fuzzy NN based method, it has been compared with a MLP NN method. Both of the methods are applied to the IEEE standard New England 10-machine 39-bus test system. The simulation results show the effectiveness of the proposed method in compare to the frequently used MLP NN based method in terms of accuracy and computational cost of CCT estimation for sample fault scenarios.

Assessing the Predictive Utility of Logistic Regression, Classification and Regression Tree, Chi-Squared Automatic Interaction Detection, and Neural Network Models in Predicting Inmate Misconduct

This study assesses the relative utility of a traditional regression approach - logistic regression (LR) - and three classification techniques - classification and regression tree (CART), chi-squared automatic interaction detection (CHAID), and multi-layer perceptron neural network (MLPNN)—in predicting inmate misconduct. The four models were tested using a sample of inmates held in state and federal prisons and predictors derived from the importation model on inmate adaptation. Multi-validation procedure and multiple evaluation indicators were used to evaluate and report the predictive accuracy. The overall accuracy of the four models varied between 0.60 and 0.66 with an overall AUC range of 0.60–0.70. The LR and MLPNN methods performed significantly better than the CART and CHAID techniques at identifying misbehaving inmates and the CHAID method outperformed the CART approach in classifying defied inmates. The MLPNN method performed significantly better than the LR technique in predicting inmate misconduct among the training samples.

The prediction of moisture through the use of neural networks MLP type

Neural networks are powerful techniques for non-linear data, which have been proven in many domains. As a result, the artificial neural networks have been applied in various fields including meteorology and climatology. This work is a contribution to the development of methods of weather prediction in general, and humidity rate in particular. In a first step, methods that are based on the study of artificial neural networks types MLP (Multilayer Perceptron) are applied for the prediction of moisture in the area of Chefchaouen in Morocco. In a second step, the proposed new architecture of neural networks of MLP type was compared to the model of multiple linear regression (MLR). The basis of learning neural model was collected between 2008 and 2011 during 1248 days. The latter consists of a number of climatic parameters, such as atmospheric pressure, air temperature, visibility, cloud cover, precipitation, dew point temperature, wind speed and humidity.Predictive models established by the MLP neural networks method are more powerful compared to those established by multiple linear regression, because of the fact that good correlation was obtained with the parameters from a neural approach with a mean squared error 5%.

A method for ranking compounds based on their relative toxicity using neural networking, C. elegans, axenic liquid culture, and the COPAS parameters TOF and EXT

Caenorhabditis elegans (L1s) were exposed to (in order of decreasing toxicity) sodium arsenite, sodium fluoride, caffeine, valproic acid, sodium borate, or dimethyl sulfoxide in C. elegans habitation medium (CeHM) for 72 consecutive hours. At this time point nematode growth and development were assessed using a Complex Object Parametric Analyzer and Sorter (COPAS ™ ). The COPAS generated biomarkers of growth (time of flight (TOF) - a measure of axial length) and development (extinction (EXT) - a measure of optical density) were subse- quently utilized to rank compounds according to their relative toxicity, as measured by the rat oral LD-50, using artificial neural network methods. Neural network methods were utilized to analyze this data because of their ability to model nonlinear endpoints and a multilayer perceptron neural network method was used because of its capability to function well in the presence of collinearity. Using a neural network approach we found that the LD-50 was correctly predicted 96% of the time. The present study demonstrates that neural network methods can be utilized to rank compounds according to their relative toxicity using COPAS-generated data (TOF and EXT) obtained from exposing a large number of nematodes to water-soluble compounds