Multilayer Perceptron (MLP)

Introduction
Snow monitoring and estimation of runoff from snow melting play an important role in controlling and managing watersheds and reservoirs and flood warning systems more efficiently. Given that the Koohrang area is found to be one of the snowy mountains in the country, and thanks to the high volume of runoff from precipitation, and snow melt, it plays a vital role in water supply for drinking, industry, and agriculture, neighboring provinces and even Iran as a whole so that it meets 10 % of total water demands in Iran. Accurate estimation of runoff from snowmelt entails spatial distribution of snow so that spatial variability of snow depth is measured via measuring snow depth in close resolution. On the other hand, the non-availability of gauge stations and extreme sampling conditions in snowy watersheds have caused researchers to think of simple and indirect strategies including regression techniques, interpolation methods, artificial intelligence, data mining, and also the use of satellite images, especially the use of radar interferometric method. Given the importance of snow depth variations and accurate estimation, although many methods have been used, there is an urgent need formore precise calculation and strategic position in this area requires procedures that are more accurate and more effective variables that are used in snow depth estimation. Study of artificial intelligence techniques and linear regression analysis and principal component analysis (PCA) along with geomorphometry parameters and inputs as well as satellite images were used to estimate the snow depth he and the results were compared. Therefore, in this study, unlike previous studies used much more variables to model snow depth, and also, the digital elevation model with the higher spatial resolution was used to model snow depth in a more accurate manner..

Methods and Materials
Koohrang region is located in the west and Chaharmahal and Bakhtiari Province with an area of over 3700 km2. It is characterized by unique climatology, hydrology, and topography. Climatic characteristics of the region include an average annual temperature of 8.5 C, rainfall of 1430 mm, a frost period of 130 days, and a winter rainfall regime. In this research, using the hypercube technique, first, 100 points were selected for sampling in the Chalgerd area. In addition to these points, 195 other points were randomly collected from the study area. To obtain the data required for this research in field work, sampling was done over three days by the Monte-Rose model sampler. After the collection of snow samples, auxiliary data required for zoning, which includes data related to satellite images and variables derived from the digital elevation model, was extracted in the Saga software environment. The artificial neural network (ANN) was chosen as a new computing system and method to estimate snow depth using morphometric and climatic information related to snow depth. After extraction of the auxiliary variables in the study, between 32 input variables and snow depth, multiple linear regression analysis was conducted to test this model is 295 points. In order to fit the multiple regression equations, snow depth data as the dependent variable and physical variables as independent variables were considered. After obtaining an equation relating to the model was tested on regression test data (20% of data) to determine the accuracy of the model to predict the snow depth. In this study, in order to reduce the number of input data to the ANN and linear regression models, the PCA method was used, and finally, the number of components was chosen to be eight. For model evaluation, the predicted snow depth was evaluated using a linear regression model and ANN followed by calculating RMSE and R2.

Results and Discussion
By trial and error, we found that a multi-layer neural network with a sigmoid activation function and a hidden layer of snow 1-6-32 for the optimal structure for the network as well as the number of repetitions and the coefficient of torque and 0.7 and 1000 was found. To evaluate and compare the performance of ANN, test data (20%) were used. ANN output values were compared with the corresponding observational values and details on the correlation coefficient were extracted. So as can be seen in the results, ANN and regression accounted for snow depth variation of 62 and 46% respectively and this regression model was significant at a probability level of 5%. The results of the PCA are to reduce the number of entries after the model of 32 to 8, the values in the model ANN and linear regression coefficient was reduced and root mean square error (RMSE) increases, and the 55 and 45 % variations in snow depth have been able to properly modeled. The less R2 and RMSE, the more accurate model is. Thus, according to the error criteria value, the ANN model outperforms other ones. According to the results obtained of all variables used in ANN, the most important variables affecting the spatial variability of snow depth in the study area in order of importance, include profile longitudinal curvature, general curvature, gradient, transverse profile curvature, watershed gradient, slope middle position, wind, normalized elevation, geographical directions, and snow normalized difference index. It is worth noting that additional variables with negligible contributions were neglected. Given that prevalent winds blow in west and southwest directions and most of the highlands are nestled in these directions, much more snow accumulation can be found in this direction than those north, east and southward directions.

Conclusion
In the present research, to estimate the spatial distribution of the snow, the four models of ANNs, linear regression, PCA, and neural network were considered. After reviewing the methods according to the statistical criteria, the lowest error rate was attributed to ANN (RMSE, 19.57), followed by PCA using ANN (RMSE, 20.86), then linear regression (RMSE, 21.09), and the highest error rate on PCA using linear regression (RMSE, 21.59). Of all variables used in ANN, the most important variables affecting the spatial variability of snow depth in the study area in order of importance, include profile longitudinal curvature, general curvature, gradient, transverse profile curvature, watershed gradient, slope middle position, wind, normalized elevation, geographical directions, and snow normalized difference index. Therefore, digital elevation models with different resolutions in modeling can be used. However, here, variables such as vegetation, geology, solar radiation were not used and therefore it is recommended to use these variables in similar studies and different time resolutions. However, in future research, the most effective variables mentioned here can be promising for accurate zonation of snow depth in snowy watersheds.

The land of Iran is located on the dry belt of the world, and despite the scarcity of water, it has been constantly exposed to the fluctuations and irregularities of the rainfall regime, and this issue has created many problems and dilemmas for the management of the country's water resources, which are considered important. It has provided the most sources of water supply needed for agriculture, drinking and industry. Therefore, studying and evaluating the quality of these sources is more important than ever. In this regard, based on the FAO Penman-Mantith relationship, the monthly potential evaporation-transpiration rate in the selected synoptic stations was calculated from the monthly meteorological data and as an input to the hybrid meta-exploratory models including the artificial neural network (ANN). enhanced support vector (ESVM), meta combinatorial learning model (Stacking) is used.

The development of information technology is currently growing very rapidly, including the impact on the hardware used. This can be exemplified in the use of hard drives that are starting to switch to SSDs. The process of selecting an SSD product to be used cannot be separated from the sources of information found on the internet. Through the internet, every user can provide reviews, both positive and negative reviews. With the many reviews regarding the review of the Samsung 850 Evo SSD on the NewEgg Store, the author uses it to be processed into information, which will have new knowledge. Based on that, the author makes research, in the form of opinion classification by analyzing sentiment through a text mining approach. In this study, two classification models were used, namely Decision Tree and Support Vector Machine. The results of this study are in the form of a comparison of the 2 models used based on the accuracy and AUC values. Based on research, the Support Vector Machine ...

This study studies the usefulness of different algorithms based on machine learning viz. Logistic Regression and Decision Tree, in forecasting the real-time ad clicks. Python has been used to implement and evaluate these models on a static dataset. Compared to Logistic Regression, the Decision Tree model has established better performance i.e. attaining an F1-score of 0.744 versus 0.7154. However, the Decision Tree's performance might be affected by overfitting and dataset characteristics. Future research should explore joint methods, deep learning, addressing imbalanced data, and privacy-preserving techniques to improve ad click prediction models.

Cancer in breasts appears as a terrible malediction in society. It snitches huge human lives across the world and its peril is going to increase at a startling rate. Identification of this disease at the initial stages is indispensable. In many cases, traditional methods are prone to errors and protracted. Models applying machine learning approaches have been shown fruitful in this application area. There are large numbers of approaches in machine learning which demonstrate impressive results. This research strives to take out the short comings from the existing models and, by resolving the underlying technical issues, deliver higher accuracy in end results. The research motivates and endeavours to make the patients' treatment processes more justified and cost-effective. The research works with WDBC dataset for breast cancer, which is publicly accessible from the UCI research database. This study uses multiple individual learners, namely Support Vector Machines (SVM), Logistic Regression(LR), Random Forest(RF), Naive Bayes(NB), K-Nearest Neighbours(K-NN), Decision Tree(DT) and an ensemble learner called Gradient Boosting(GB) with multiple techniques of feature selection namely Principal Component Analysis (PCA) and Recursive Feature Elimination (RFE). The experimental techniques discern subtle patterns within the dataset. The proposed model evaluates the results and performances through metrics specificity, sensitivity and accuracy in a comparative structure. It succeeds with higher accuracy of 98%. The study highlights its potential as a significant tool in medical diagnostics.

Heart failure (HF) is a common complication of cardiovascular diseases. This research focuses on assessing the effectiveness of different models for predicting HF using both Traditional Machine Learning (TML) methods and Automated Machine Learning (AutoML) approaches. TML models need extensive manual tuning and expert knowledge for algorithm selection and optimization, making the process slow and susceptible to human error. To tackle this challenge, the work proposed an AutoML approach utilizing the AutoGluon framework for predicting HF. The main goal of this study is to automate the process of selecting the most efficient model. This study compares a total of twenty (20) individual-trained ML models, consisting of fourteen (14) from AutoML and six (6) from TML. In TML, Logistic Regression (LR) produced the highest 87.50% accuracy and ROC-AUC of 88.83% compared to Support Vector Models (SVM), Decision Trees (DT), Gaussian Naïve Bayes (GNB), Random Forests (RF) and K-Nearest Neighbors (KNN). In AutoML, the CatBoost model outperforms the other thirteen algorithms with the highest accuracy of 99.39% and ROC-AUC of 99.89%. The results show that an AutoML based algorithm called the CatBoost model gives the most accurate model among all 20 models. SHAP was employed to interpret the top-performing model, increasing its transparency and usability.

Artificial Neural Networks (ANN) demonstrate a compelling application of AI in predicting student performance, a critical aspect for both students and educators. Accurate forecasting of student achievements enables educators to monitor progress effectively, allowing educational institutions to optimize outcomes and improve student results. This study focuses on leveraging ANN for predictive analytics in student performance. Through a detailed evaluation of transfer functions, optimizers, learning rates, and momentum values, the model achieves an impressive 98% accuracy with specific configurations: a learning rate of 0.005, momentum of 0.7, Sigmoid transfer function, and SGD optimizer. Additionally, the study performs a comparative analysis of various Machine Learning Algorithms, including ANN, Support Vector Machines (SVM), K-Nearest Neighbors (KNN), Decision Trees (DT), Naive Bayes (NB), and Logistic Regression (LR). Using data from 689 B.Tech students at IP University, the analysis reveals that ANN outperforms other algorithms with an accuracy of 97%. This high accuracy demonstrates the potential of ANN in educational settings, providing a valuable tool for educators to enhance student performance and outcomes.

This article appeared in a journal published by Elsevier. The attached copy is furnished to the author for internal non-commercial research and education use, including for instruction at the authors institution and sharing with colleagues. Other uses, including reproduction and distribution, or selling or licensing copies, or posting to personal, institutional or third party websites are prohibited. In most cases authors are permitted to post their version of the article (e.g. in Word or Tex form) to their personal website or institutional repository. Authors requiring further information regarding Elsevier's archiving and manuscript policies are encouraged to visit: http://www.elsevier.com/copyright

In acoustic engineering, the types of material used in a room are basically one of the fundamental features that are essential in some of room acoustic parameters computation. This paper proposed an improved system to identify room material type from its surface photographic image. Data images of several room surfaces were collected for the system input. This improved system implements Gray Level Co-occurrence Matrix (GLCM) and modified Zernike moments for image extraction and hybrid Particle Swarm Optimization and backpropagation (PSO-BP) algorithm for classification. For comparison purpose, experiments using variations combination of GLCM and modified Zernike moments extraction as well as Levenberg-Marquardt, back-propagation neural network (BPNN), and PSO-BP algorithm were executed. By applying the proposed methods, the system accuracy increased around 30% compared to previous research. Moreover, the convergence attained during training was three times faster compared to BP algorithm. Thus using the new methods in identifying material surface images had positively improved the system in becoming more efficient and reliable.

Time series are an important object of study in sciences, engineering and business, especially in cases where it is expected to know, predict and optimize behaviors. In this context, we intend to show the feasibility of using artificial neural networks in the study of several time series in an engineering course, especially those that have no overt behavior or are not able to be modeled mathematically in a simple way and have direct application in the education of future engineers.

In this paper, we propose a new modified particle swarm algorithm for training some neural network classifiers for the most used classification problems in literature. Researches in artificial neural network field are based on different network architectures including multilayer perceptron, single multiplicative neuron and pi-sigma neuron model. To obtain a satisfactory performance for these classifiers, one of the most important issues is network training. Evolutionary algorithms are commonly used for training neural network classifiers. Particle swarm algorithm is population based, stochastic and meta-heuristic algorithm to solve optimization problems. As all evolutionary algorithms, the particle swarm algorithm may fall into local optimum and convergence rate may incredibly decline in iterative process. To overcome these shortcomings we refer to modify the particle swarm optimization with changing position matrix for each generation at iterative process. Experimental results show...

With recent advances in Electroencephalogram (EEG) signal processing and biomedical instrumentation, brain machine interfaces are used for rehabilitation of people suffering from neuromuscular disorders. This paper presents a novel method employing Hierarchical classifier using optimised Neural Networks to classify left-hand movement, right-hand movement and word generation using EEG signals. One of the most important components of brain computer interface (BCI) is feature extraction of EEG signals. Power spectral density (PSD) is used for feature extraction from EEG signals. The proposed pre-processing and reconfiguration of PSD samples make them more discriminative & yield appropriately organized feature vectors. The adaptation of network weights using Comprehensive Learning Particle Swarm Optimization (CLPSO) is proposed to improve the performance of Neural Network (NN). Further, the two level hierarchical neural network is used to enhance the discriminative property of the featu...

In this paper, we propose a new modified particle swarm algorithm for training some neural network classifiers for the most used classification problems in literature. Researches in artificial neural network field are based on different network architectures including multilayer perceptron, single multiplicative neuron and pi-sigma neuron model. To obtain a satisfactory performance for these classifiers, one of the most important issues is network training. Evolutionary algorithms are commonly used for training neural network classifiers. Particle swarm algorithm is population based, stochastic and meta-heuristic algorithm to solve optimization problems. As all evolutionary algorithms, the particle swarm algorithm may fall into local optimum and convergence rate may incredibly decline in iterative process. To overcome these shortcomings we refer to modify the particle swarm optimization with changing position matrix for each generation at iterative process. Experimental results show...

Results and discussion The results of Pearson's correlation coefficient showed that out of the eight considered parameters, cloudiness parameters, average maximum temperature, water vapor pressure, maximum relative humidity, and dew point have the most significant correlation at the 99% level with rainfall, so parameters The inputs were considered to predict the rainfall parameter in the SVM-SA algorithm. Based on the results, using five non-precipitation meteorological parameters, including cloud cover, maximum temperature, water vapor pressure, maximum relative humidity, and dew point, the RMSE, SE, and R 2 in the training section are equal to 6.02 mm, 0.01, and 0.999. In the testing section, 18.72 mm, 0.03, and 0.925 mm, were calculated, respectively. Conclusions The results showed that the SVM-SA could be highly accurate in simulating precipitation changes in the study area, and this model can be introduced as a new hybrid algorithm in engineering sciences.

The dependable operation of brain-computer interfaces (BCI) based on electro electroencephalogram (EEG) signals requires precise classification of multi-channel EEG signals. The design of EEG interpretation and classifiers for BCI are open research questions whose difficulty stems from the need to extract complex spatial and temporal patterns from noisy multidimensional time series obtained from EEG measurements. In this paper we attempt to classify EEG data used in the BCI competition by the combination of pattern classification methods. We use Common Spatial Pattern (CSP) to extract features. A Genetic Algorithm (GA) was applied first to evolve an artificial neural network (ANN) to find the optimum structure of ANN. A Particle Swarm Optimization (PSO) was also attempted to determine the optimal number of hidden neurons complementary to the GA approach. Then the GA was used to evolve the connection weights of the ANN. I. INTRODUCTION rain-Computer Interface is a direct communication channel between brain and computer or external devices. They create a link between the brain and an output device by bypassing conventional motor output pathways of nerves and muscles. The translation of user intent into device control commands is a major challenge. Success requires the effective interaction of two adaptive controllers: the user's brain, which produces brain activity that encodes intent, and the BCI system, which translates that activity into device control commands. In the literature, many machine learning and pattern classification algorithms have been reported to give excellent results when applied to BCI data in offline analyses. A typical noninvasive electroencephalogram (EEG) based brain-computer communication devices are composed of three subsystems. Namely: EEG acquisition, EEG signal processing and the output subsystems. The acquired EEG signal can be regarded a complex time series signal that has

Time series are an important object of study in sciences, engineering and business, especially in cases where it is expected to know, predict and optimize behaviors. In this context, we intend to show the feasibility of using artificial neural networks in the study of several time series in an engineering course, especially those that have no overt behavior or are not able to be modeled mathematically in a simple way and have direct application in the education of future engineers.

Electroencephalogram (EEG) is used routinely for diagnosis of diseases occurring in the brain. It is a very useful clinical tool in the classification of epileptic seizures and the diagnosis of epilepsy. In this study, epilepsy diagnosis has been investigated using EEG records. For this purpose, an artificial neural network (ANN), widely used and known as an active classification technique, is applied. The particle swarm optimization (PSO) method, which does not need gradient calculation, derivative information, or any solution of differential equations, is preferred as the training algorithm for the ANN. A PSO-based neural network (PSONN) model is diversified according to PSO versions, and 7 PSO-based neural network models are described. Among these models, PSONN3 and PSONN4 are determined to be appropriate models for epilepsy diagnosis due to having better classification accuracy. The training methods-based PSO versions are compared with the backpropagation algorithm, which is a traditional method. In addition, different numbers of neurons, iterations/generations, and swarm sizes have been considered and tried. Results obtained from the models are evaluated, interpreted, and compared with the results of earlier works done with the same dataset in the literature.

This paper presents the results obtained by the experiments carried out in the project which aims to classify EEG signal for motor imagery into right hand movement and left hand movement in Brain Computer Interface (BCI) applications. In this project the feature extraction of the EEG signal has been carried out using Discrete Wavelet Transform (DWT). The wavelet coefficients as features has been classified using Support Vector Machine (SVM), k-Nearest Neighbor (k-NN) and Backpropagation Neural Network (BPNN). The maximum classification accuracy obtained using SVM is 78.57%, using k-NN is 72% and using BPNN is 80%.

This research explains the application of ElectroEncephaloGraph (EEG) signal waves used to move the up cursor and down cursor. In each sub band of the waveform, Electro Encephalo Graph (EEG) will produce the average and standard deviation to be used as a feature of the EEG. Artificial Neural Network Backpropagation as the basis for determining whether the cursor moves up or the cursor moves down. The data used in this study is EEG data derived from BCI Competition 2003 (BCI Competition 2003). Decision-making is done in two stages. In the first stage, the mean and standard deviation values on each wavelet subband as a feature extraction of EEG data. This feature is an input to the Backpropagation Neural Network. In the second stage of the process of inserting into two classes (class 0 and class 1) EEG data files, there are 260 EEG and 293 file training data files from EEG file test data files, totaling to 553 EEG data files. The results obtained for the classification of EEG results ...

Through the development of technologies such as control systems, high-performance processor and more durable batteries, the investigations on the design of stand-alone limb prosthesis are more achievable. As a result, designing rehabilitation devices for handicapped people opens special field of study for researches. Prosthetic hands stand out in this research because of the functionality in everyday life. Hand movements classifications using EMG signals based on Artificial Neural Network (ANN) has been used very effectively. In this paper, two ANN classification algorithms were developed that allow motion classification for prosthetic hands. Multilayer Perceptron (MLP) and radial basis neural network (RBF) architectures were used and their performance was compared. For this purpose, "holding pen" and "gripping" movements were classified. The experimental results showed that MLP structure has higher performance than RBF structure for these movements.

Stock market Indices prediction is one of the most important issues in the financial field. Although many prediction models have been developed during the last decade, they suffer a poor performance because indices movement is highly non stationary and volati le dynamic process. Artificial Neural Network (ANN) is a technique that is heavily researched and widely used in stock market prediction [1][6]. However, there is no formal method to determine the optimal neural network for prediction purpose in the literature. This paper describes various Neural Network models for stock prediction. The prediction was done by, Modular Neural Network, ARIMA-based Neural Networks, Genetic Algorithm, Recurrent Network, Back propagation Network, Radial Basis Function, Branch Network, Functional Link Artificial Neural Network, Feed Forward Neural Network, Fuzzy Neural Network etc [8]. Analysis of all these Neural Network models is performed in this paper, as well as the future work.

A Brain-Computer Interface (BCI), also known as Brain Machine Interface (BMI), is a communication system that lets the users to interact with electronic devices by means of control signals acquired from Electroencephalographic (EEG) activity without engaging peripheral nerves and muscles. The preliminary motivation for BCI research was to develop assistive devices for people with locked-in disabilities. Nowadays, researchers are exploring BCI as a novel anthropomorphic interaction channel for daily applications such as robotics, virtual reality, and games. This paper investigates the effect of weight optimization using Hybrid Particle Swarm Optimization (PSO) for Multilayer Perceptron Neural Network classifier which uses features selected by Principal Component Analysis and Hybrid PSO.

Time series are an important object of study in sciences, engineering and business, especially in cases where it is expected to know, predict and optimize behaviors. In this context, we intend to show the feasibility of using artificial neural networks in the study of several time series in an engineering course, especially those that have no overt behavior or are not able to be modeled mathematically in a simple way and have direct application in the education of future engineers.

Brain-Computer Interfaces (BCIs) measure brain signals activity, intentionally and unintentionally induced by users, and provides a communication channel without depending on the brain's normal peripheral nerves and muscles output pathway. Feature Selection (FS) is a global optimization machine learning problem that reduces features, removes irrelevant and noisy data resulting in acceptable recognition accuracy. It is a vital step affecting pattern recognition system performance. This study presents a new Binary Particle Swarm Optimization (BPSO) based feature selection algorithm. Multi-layer Perceptron Neural Network (MLPNN) classifier with backpropagation training algorithm and Levenberg-Marquardt training algorithm classify selected features.

With the advancement of Machine Learning, since its beginning and over the last years, a special attention has been given to the Artificial Neural Network. As an inspiration from natural selection of animal groups and human’s neural system, the Artificial Neural Network also known as Neural Networks has become the new computational power which is used for solving real world problems. Neural Networks alone as a concept involve various methods for achieving their success; thus, this review paper describes an overview of such methods called Particle Swarm Optimization, Backpropagation, and Neural Network itself, respectively. A brief explanation of the concepts, history, performances, advantages, and disadvantages is given, followed by the latest researches done on these methods. A description of solutions and applications on various industrial sectors such as Medicine or Information Technology has been provided. The last part briefly discusses the directions, current, and future chall...

Background and Objectives: Temperature and rainfall are two important meteorological variables, especially in arid and semi-arid areas. As a result, determining the value of these variables, their changes and prediction of these phenomena are necessary for more precise planning in the management of agricultural, economic and social sectors. Nowadays, incompatibility of temporal and spatial scales required in investigated models on the effect of climate change with GCM outputs and the need to assess the change trend in meteorological threshold variables at the regional scale has led to develop various downscaling methods. So, the aim of this study is the comparative comparison of data mining models in downscaling of rainfall and temperature based on data of NCEP general circulation model.
Material and Methods: The study area in this research is bazoft- e- Samsami watershed. This basin is one of the northern Karun sub-basins located in the northwest of Chaharmahal and Bakhtiari province. Marghak rain gauge and hydrometric stations are located at its outlet.
In this study, the performance and efficiency of four methods including decision tree (M5), Nearest Neighbor (KNN), Multilayer Perceptron (MLP) and Simple linear regression (SLR) were evaluated for modeling monthly rainfall and temperature of Marghak station during the training period of 1971-1990 and The 1991-2000 test period using NCEP output parameters.
Results: Monthly rainfall modeling results using mentioned models showed that the output of all models except the KNN model provides negative values for rainfall. The rainfall prediction by M5 model in January, March, April and December is lower than the observed values (P). This situation is also somewhat seen in other models. Also, given that the minimum rainfall is zero, it can be concluded from the low predicted values rather than observed values that the maximum limit of rainfall with these models is not well predicted. The prediction of rainfall by all models in all months except May has a lower standard deviation than the observed values (P).
The predicted results of monthly temperature also showed that only MLP output provides negative values for the temperature, which can be due to the extrapolation and generalizationin in MLP method. Also, The standard deviation obtained from all models in January, February, March, April, July, August, October, November and December is more than standard deviation of observed temperature. The results of statistical analyzes also showed that M5 than the other models in the test stage according to RMSE, MBE and R2 have better estimates for rainfall and monthly temperature. Although the results of determination coefficient (R2) in the test stage for monthly temperature estimation are weaker than monthly rainfall.
Conclusion:
The results of the efficiency of four models of KNN, M5, SLR and MLP in monthly rainfall and temperature modeling in Marghak meteorological station with NCEP output data showed that these models were weak in downscaling the monthly rainfall and temperature. Therefore, despite the relative superiority of M5 model compared to other models, the use of these data mining models is not recommended to predict rainfall and temperature variables in Margak station.

The dependable operation of brain-computer interfaces (BCI) based on electroencephalogram (EEG) signals requires precise classification of multi-channel EEG signals. The design of EEG interpretation and classifiers for BCI are open research questions whose difficulty stems from the need to extract complex spatial and temporal patterns from noisy multidimensional time series obtained from EEG measurements. In this paper we attempt to

Electroencephalogram (EEG) is the most significant signal for brain-computer interfaces (BCI). Nowadays, motor imagery (MI) movement based BCI is highly accepted method for. This paper proposes a novel method based on the combined utilization of principal component analysis (PCA), wavelet packet transformation (WPT), and two-stage machine learning algorithm to classify four-class MI EEG signal. This work includes four-class MI events by an imaginary lifting of the left hand, right hand, left foot, and Right Foot. The main challenge of this work is to discriminate the similar lobe EEG signal pattern such as left foot VS left hand. Another critical problem is to identify the MI movements of two different feet because their activation level is very low and show an almost similar pattern. This work firstly uses the PCA to reduce the signal dimensions of the left and right lobe of the brain. Then, WPT is used to extract the feature from the different class EEG signal. Finally, the artificial neural network is trained into two stages-1 st stage identifies the lobe from the signal pattern and the 2 nd stage identifies whether the signal is of MI hand or MI foot movement. The proposed method is applied to the 4-class MI movement related EEG signals of 15 participants and found excellent classification accuracy (>74% on average). The outcomes of the proposed method prove its effectiveness in practical BCI implementation.

The outbreak of COVID-19 has brought the world to an unprecedented position where financial and mental resources are drying up. Livelihoods are being lost, and it is becoming tough to save lives. These are the times to think of unprecedented solutions to the financial challenges being faced. Artificial intelligence (AI) has provided a fresh approach to finance through its implementation in the prediction of financial market prices by promising more generalizable results for stock market forecasting. Immense literature has attempted to apply AI and machine learning for predicting stock market returns and volatilities. The research on the applications of AI in finance lacks a consolidated overview of different research directions, findings, methodological approaches, and contributions. Therefore, there is a need to consolidate the extant literature in this upcoming field to consolidate the findings, identify the research gaps in the existing literature, and set a research agenda for f...

Brain-computer interfaces (BCI) are devices that enable communication between a computer and humans by using brain activity as input signals. Brain imaging technology used in a BCI system is usually electroencephalography (EEG). In order to properly interpret brain activity, acquired signals from the brain have to be classified correctly. In this paper EEG signals are transformed by means of discrete wavelet transform. Thus the obtained signal features are used as inputs for a neural network classifier that should separate five different sets of EEG signals representing various mental tasks. Mean classification accuracy for the recognition of all five tasks was 90.75% and mean classification accuracy for the recognition of two tasks (baseline and any other mental task) was 99.87%. The same procedure was also used on the motor imagery dataset. A mean classification accuracy of 68.21% suggests alternative methods of feature extraction for motor imagery tasks.

In acoustic engineering, the types of material used in a room are basically one of the fundamental features that are essential in some of room acoustic parameters computation. This paper proposed an improved system to identify room material type from its surface photographic image. Data images of several room surfaces were collected for the system input. This improved system implements Gray Level Co-occurrence Matrix (GLCM) and modified Zernike moments for image extraction and hybrid Particle Swarm Optimization and back-propagation (PSO-BP) algorithm for classification. For comparison purpose, experiments using variations combination of GLCM and modified Zernike moments extraction as well as Levenberg-Marquardt, back-propagation neural network (BPNN), and PSO-BP algorithm were executed. By applying the proposed methods, the system accuracy increased around 30% compared to previous research. Moreover, the convergence attained during training was three times faster compared to BP algo...

Image segmentation can be posed as a multiclass classification problem. In doing so, segmentation evaluation can be made through multiclass classification errors. Instead of being used for evaluation, in this work the mean multiclass type I and II errors are proposed for multilayer perceptron training via PSO. Moreover, some relations involving mean multiclass errors and conditional errors are exposed. Applied to image segmentation, mean multiclass errors were compared to mean squared error (MSE) as objective functions. The approach was effective and able to provide accuracy and precision gains, resulting in a lower number of function evaluations in a cross-validated experiment.

e outbreak of COVID-19 has brought the world to an unprecedented position where financial and mental resources are drying up. Livelihoods are being lost, and it is becoming tough to save lives. ese are the times to think of unprecedented solutions to the financial challenges being faced. Artificial intelligence (AI) has provided a fresh approach to finance through its implementation in the prediction of financial market prices by promising more generalizable results for stock market forecasting. Immense literature has attempted to apply AI and machine learning for predicting stock market returns and volatilities. e research on the applications of AI in finance lacks a consolidated overview of different research directions, findings, methodological approaches, and contributions. erefore, there is a need to consolidate the extant literature in this upcoming field to consolidate the findings, identify the research gaps in the existing literature, and set a research agenda for future researchers. is paper addresses this need by synthesizing the extant literature in the form of a systematic review for addressing the use of AI in stock market predictions and interpreting the results in a narrative review. e gap formed through this article is the use of a combination of AI as a subject with the neural network as another area and stock market forecasting as another theme, and it will pave the way for future research studies. e analyses help highlight four important gaps in the existing literature on the subject.

https://www.ijert.org/enhancement-of-finger-vein-and-low-resolution-fingerprint-images-for-authentication

https://www.ijert.org/research/enhancement-of-finger-vein-and-low-resolution-fingerprint-images-for-authentication-IJERTV3IS030725.pdf

Biometrics is mainly used for human identification by using different physical traits. The traits that can be used as biometrics are face, palm print, voice, signature, gait etc. But use of these traits in biometrics is not perfectly reliable or secure. So, inorder to overcome the security issue, a non-forgeable pattern has to be used. The finger-vein is a promising biometric pattern for personal identification in terms of its security and convenience. The fingerprints are also unique like finger-vein pattern. So it can be used along with finger-vein as biometric trait. In this paper, the images are enhanced by incorporating the concept of local histogram equalization, which improves the local contrast of an image.

Multimodal system is capable of increasing the scope and variety of input information the system takes from the users for authentication. However, Face, Ear and fingerprint have compatibility formation but little research have been done in the area of comparing biometric features in order to determine their overall accuracy in multimodal systems. A total of 2160 datasets were used for this experiment. 1260 datasets were used for the training and 900 were used for testing. These images were preprocessed using histogram equalization and feature extraction was carried out using Principal component analysis (PCA). Self-organizing feature map (SOFM) and back propagation neural network (BPNN) was used for classification. The performance of the developed multimodal biometric systems (face, ear and finger) was compared and evaluated in MATLAB environment. The results showed that SOFM has high recognition accuracy and time than BPNN.

Nowadays, Fingerprint recognition is the one of the authentication used for security applications. It provides hope for the society in reliable authentic biometric systems. Fingerprint technology emerges in various sectors such as government, organizations, libraries, universities, banks etc. It is widely used for biometric systems other than Iris, Face, Hand, Voice and Signature because of its uniqueness and distinctness. Traditional methods are not effectively used for analyzing the texture feature of finger print than neural network classifier. Fingerprint recognition will be identified and classified with the help of Euclidean distance and NN classifier for better accuracy has proposed in this paper. It uses certain techniques in preprocessing the image such as Histogram equalization and Fast Fourier transform. The performance of the proposed approach has significant result than the existing techniques used in the finger print recognition system.

This paper presents the dynamic model identification algorithm of the continuous stirred tank reactor (CSTR) using a multi-layer perceptron (MLP) neural network topology. The neural network approach for (CSTR) dynamic modeling is trained by using a particle swarm optimization (PSO) technique as a simple and fast training unsupervised algorithm. Polywog wavelet activation function is used in the structure of MLP neural network. The identification algorithm given in this paper has been proved to be reasonable and precise via Matlab simulation results in terms of fast, stable and minimum number of fitness evaluation for the CSTR modeling. Coolant C af , q , T f Reactant q c (t) , T c (t) C a (t) , q , T(t) Product Figure 1: CSTR with cooling jacket [4,5].

The world is changing, and this is the digital era. Almost everything around us is digitized and the flow of information is huge from a variety of sources ranging from mobile phone, smart devices, surveillance, sensors of the universe, weather forecasting sensors, medical equipment, customers transactions of the internet, user behaviours on the internet, and so on. Billions of dollars get wasted every year due to fraud. Traditional methods of fraud detection play an important role in minimizing these losses. Increasingly fraudsters have developed a variety of way to elude their detection, both by working together and by leveraging various other means of constructing fake identities. This paper proposed a new approach for fraud prevention in different sector with help of graph database by identifying of previous fraud records in graph database.

The rainfall-runoff relationship is one of the most complex hydrological phenomena. In recent years, hydrologists have successfully applied backpropagation neural network as a tool to model various nonlinear hydrological processes because of its ability to generalize patterns in imprecise or noisy and ambiguous input and output data sets. However, the backpropagation neural network convergence rate is relatively slow and solutions can be trapped at local minima. Hence, in this study, a new evolutionary algorithm, namely, particle swarm optimization is proposed to train the feedforward neural network. This particle swarm optimization feedforward neural network is applied to model the daily rainfall-runoff relationship in Sungai Bedup Basin, Sarawak, Malaysia. The model performance is measured using the coefficient of correlation and the Nash-Sutcliffe coefficient. The input data to the model are current rainfall, antecedent rainfall and antecedent runoff, while the output is current runoff. Particle swarm optimization feedforward neural network simulated the current runoff accurately with R = 0.872 and E 2 = 0.775 for the training data set and R = 0.900 and E 2 = 0.807 for testing data set. Thus, it can be concluded that the particle swarm optimization feedforward neural network method can be successfully used to model the rainfall-runoff relationship in Bedup Basin and it could be to be applied to other basins.

Brain Computer Interface (BCI) research is based on recording and analyzing electroencephalographic (EEG) data and recognizing EEG patterns associated with various mental states. BCIs had become an active research area in the last decade. The ability of The Multilayer Perceptron Neural Networks (MLPNN) to model non-linear relationship is appropriate to model the complex nature of EEG signals. MLPNN’s convergence rate is relatively slow and it often yields sub-optimal solutions. In this paper, it is proposed to enhance the classification ability of the MLPNN incorporating Particle Swarm Optimization (PSO). This optimization helps to overcome the slow convergence rate and to avoid local minimum. The features of motor imagery in the frequency domain are extracted using Hilbert transform, and Principal Component Analysis is used for feature reduction. This is classified by MLPNN modified with PSO.

In this report, two learning methods of a multilayer perceptron neural network have been studied. We apply Levenberg-Marquardt and error backpropagation (EBP) methods on the vision system of Small Size League (SSL) robots to find their directions on the field. The results are compared over different training sets.

Biometric recognition systems have advanced significantly in the last decade and their use in specific applications will increase in the near future. The ability to conduct meaningful comparisons and assessments will be crucial to successful deployment and increasing biometric adoption. The best modality used as unimodal biometric systems are unable to fully address the problem of higher recognition rate. Multimodal biometric systems are  able  to  mitigate  some  of  the  limitations  encountered  in unimodal  biometric  systems,  such  as  non-universality, distinctiveness, non-acceptability, noisy sensor data, spoof attacks, and performance. More reliable recognition accuracy and performance are achievable as different modalities were being combined together and different algorithms or techniques were being used. The work presented  in  this  paper  focuses  on  a  bimodal  biometric  system using  face  and  fingerprint.  An  image  enhancement  technique (histogram  equalization)  is  used  to  enhance  the  face  and fingerprint  images.  Salient features of the face and fingerprint were extracted using the Gabor filter technique. A dimensionality reduction technique was carried out on both images extracted features using a principal component analysis technique. A feature level fusion algorithm (Mahalanobis distance technique) is used to combine each unimodal feature together. The performance of the proposed approach is validated and is effective.

Biometric recognition systems have advanced significantly in the last decade and their use in specific applications will increase in the near future. The ability to conduct meaningful comparisons and assessments will be crucial to successful deployment and increasing biometric adoption. Even the best modality and unimodal biometric systems were unable to fully address the problem of accuracy and performance in terms of their false accept rate (FAR) and false reject rate (FRR). Although multimodal biometric systems were able to mitigate some of the limitations encountered in unimodal biometric systems, such as non-universality, distinctiveness, non-acceptability, noisy sensor data, spoof attacks, and performance, the issue of low accuracy and performance still persists. In this paper, we review research papers focused on the accuracy and performance enhancement in information fusion of face and fingerprint biometric recognition systems, determine the main features of the selected methods, and then point out their merits and shortcomings.
Finally, we propose a novel approach in mitigating the problem of accuracy and performance of information fusion of multimodal biometric systems. This approach makes use of multilayer perceptron neural networks in training and testing of the network while also proposing the use of the most common used modalities (face and fingerprint) in biometric arena.