Performance Of ANN Research Articles

IMPLICATION OF SOM-ANN BASED CLUSTERING FOR MULTI-STATION RAINFALL-RUNOFF MODELING

In this study the performance of ANN with feed-forward neural network (FFNN) algorithm evaluated rainfall-runoff modeling in five gauging stations in Florida State. In addition, for investigating the performance of ANN in multi-station discharge prediction, self-organizing map (SOM) clustering tool employed in order to cluster the input data with similar patterns, due to the large amount of records in multiple stations. The main aim of study is to investigate capability and accuracy of ANN based methods in multi-station discharge prediction. In order to consider multiple stations effect on watershed outlet discharge, different combinations for precipitation and discharge data of all stations with antecedent values over the watershed have been taken into account. In this way, application of the representatives from each cluster led to significantly reduction in the numbers of the input variables so that the optimal ANN structure could be proposed. Therefore, ANN as a data-driven model was trained to predict daily runoff for the Peace River basin via recorded values from July 1995 to July 2011. Three scenarios conducted the aim of research; first scenario was an integrated ANN model trained by the data of rainfall and runoff at multiple stations. The second scenario was a sequential ANN model processed with upstream discharge records in addition to rainfall data as inputs and downstream discharge values as target. Finally, third scenario was a SOM-ANN model, in which rainfall and runoff data were clustered according the homogeneity of data via (SOM). The center of each cluster as the dominant component of each cluster was imposed to ANN in order to present an optimal rainfall-runoff model over the watershed. In all scenarios, different data sets at various time lags in both rainfall and stream flow data were applied as inputs in ANN- based model to predict stream flow. Results show that ANN model coupled with SOM is useful tools for forecasting multi-station discharge and precipitation event response in the watershed. Furthermore, the comparison of scenarios lead s to select the most efficient and optimal inputs to ANN which subsequently, presents the optimal multi- station rainfall-runoff model over the watershed.

Application of Support Vector Network for Power System Static Security Evaluation

This paper uses Support Vector Network (SVN) to examine whether the power system is secured under steady-state operating conditions. A system is considered operationally reliable if the load bus voltages do not fall below a certain limit and if the power flow through lines does not exceed the corresponding allowable values. SVN determines the minimum bus voltage and maximum ratio of line-flow to permissible line-flow. The input variables to the network are the active power of the load buses, power factor of the loads and the net generated powers of the generating buses. IEEE 14-bus system has been taken as an example. The proper kernel function and optimal value of C i. e. penalty parameter has been calculated. A comparison of the performance of SVN and ANN with those calculated by fast decoupled load flow is carried out. Results of the SVN closely agree with that obtained by fast decoupled load-flow and ANN in the case of proportional input vector. ANN is not suitable in the case of disproportionate input vector whereas SVN overcomes this disadvantage.

Forecasting Groundwater Level by Artificial Neural Networks as an Alternative Approach to Groundwater Modeling

Abstract The main purpose of this article is to apply feed forward back propagation neural network (FNN) to predict groundwater level of Aghili plain, which is located in southwestern Iran. An optimal design is completed for the two hidden layers with four different algorithms: descent with momentum (GDM), Levenberg Marquardt (LM), resilient back propagation (RP), and scaled conjugate gradient (SCG). The training data for ANN is obtained from observation data. Rain, evaporation, relative humidity, temperature, discharge of irrigation canal, and groundwater recharge from the plain boundary were used in input layer while future groundwater level was used as output layer. Before training, the available data were divided into three groups, according to hydrogeological characteristics of different parts of the plain surrounding each piezometer. Statistical analysis in terms of Mean-Square-Error (MSE) and correlation coefficient (R) was used to investigate the prediction performance of ANN. FFN-LM algorithm has shown best result in the present study for all three hydrogeological groups. Now, to predict water level, the t time data (October 2003 to July 2009) and t+1 time data (October 2004 to July 2010) were used as input and output respectively. The best condition of this network was achieved for each group of data. Next, with defining the new input data related to August 2010 to January 2011 groundwater level was predicted for the following year. The achieved results of ANN model in contrast with results of finite difference model showed very high accuracy of artificial neural network in predicting groundwater level.

Effect of Water Depth on the Performance of Intelligent Computing Models in Predicting Wave Transmission of Floating Pipe Breakwater

Understanding the physics of complex system plays an important role in selection of data for training intelligent computing models. Based on the physics of the wave transmission of Horizontally Interlaced Multilayer Moored Floating Pipe Breakwater, a laboratory experiment carried out at Department of Applied Mechanics, National Institute of Technology Surathkal, India, authors felt that relative depth of water ( d/ L) may have effect on the performance of intelligent computing models. In the present paper, d/ L is taken as one of the inputs to study the performance of ANN and Genetic Algorithm based Support Vector Machine Regression (GA-SVMR) model which was ignored by the authors in their previous studies. The performances of present ANN-1 and GA-SVMR-1 models are compared with the previous ANN and GA-SVMR models. The results revealed that there is a slight improvement in the performance of present ANN-1 and GA-SVMR-1 models in terms of Correlation Coefficient.

Fractional snow cover mapping from MODIS data using wavelet-artificial intelligence hybrid models

Summary This study was carried out to evaluate the wavelet-artificial intelligence hybrid models to produce fractional snow cover maps. At first, cloud cover was removed from MODIS data and cloud free images were produced. SVM-based binary classified ETM+ imagery were then used as reference maps in order to obtain train and test data for sub-pixel classification models. ANN and ANFIS-based modeling were performed using raw data (without wavelet-based preprocessing). In the next step, several mother wavelets and levels were used in order to decompose the original data to obtain wavelet coefficients. Then, the decomposed data were used for further modeling processes. ANN, ANFIS, wavelet-ANN and wavelet-ANFIS models were compared to evaluate the effect of wavelet transformation on the ability of artificial intelligence models. It was demonstrated that wavelet transformation as a preprocessing approach can significantly enhance the performance of ANN and ANFIS models. This study indicated an overall accuracy of 92.45% for wavelet-ANFIS model, 86.13% for wavelet-ANN, 72.23% for ANFIS model and 66.78% for ANN model. In fact, hybrid wavelet-artificial intelligence models can extract the characteristics of the original signals (i.e. model inputs) accurately through decomposing the non-stationary and complex signals into several stationary and simpler signals. The positive effect of fuzzification as well as wavelet transformation in the wavelet-ANFIS model was also confirmed.

Modeling of SVC Controller based on Adaptive PID Controller using Neural Networks

The Flexible AC Transmission System (FACTS) technology is a promising technology to achieve complete deregulation of power system based on power electronic devices, used to enhance the existing transmission capabilities in order to make the system flexible and independent in operation then the system will be kept within limits without affecting the stability. Complete closed-loop smooth control of voltage can be achieved using shunt connected FACTS devices. Static VAR Compensator (SVC) is one of the shunt connected devices, which can be utilized for the purpose of voltage and reactive power control in power systems. In this paper the considered structure of SVC consists of (TCR-FC) which is applied at SMIB system model, the dynamic equations for the (SMIB-SVC) model will be presented, the system equations expressed in terms of state space equations then by using MATLAB the plant of the system model will be presented under various loading conditions. A Neuro-PID controller model has been developed to improve on the response and performance of a conventional Proportional plus Integral plus Derivative (PID) controller which control the response of the plant model by developing a self-tuning/adaptive Neural-PID controller. The proposed Neuro-controller was developed using the back propagation algorithm. The ANN based PID (ANNPID) controller compared with ANN controller through MATLAB simulation results. Comparison of performance responses of ANN controller and ANN-PID controller show that ANN-PID controller has quite satisfactory generalization capability, feasibility and reliability, as well as the accuracy in the system; the superiority of the performance of ANN over PID controller is highlighted under various loading conditions.

Implementation of hybrid ANN–PSO algorithm on FPGA for harmonic estimation

Harmonic estimation is the main process in active filters for harmonic reduction. A hybrid Adaptive Neural Network–Particle Swarm Optimization (ANN–PSO) algorithm is being proposed for harmonic isolation. Originally Fourier Transformation is used to analyze a distorted wave. In order to improve the convergence rate and processing speed an Adaptive Neural Network Algorithm called Adaline has then been used. A further improvement has been provided to reduce the error and increase the fineness of harmonic isolation by combining PSO algorithm with Adaline algorithm. The inertia weight factor of PSO is combined along with the weight factor of Adaline and trained in Neural Network environment for better results. ANN–PSO provides uniform convergence with the convergence rate comparable that of Adaline algorithm. The proposed ANN–PSO algorithm is implemented on an FPGA. To validate the performance of ANN–PSO; results are compared with Adaline algorithm and presented herein.

An Intelligent Computing Model for Wind Speed Prediction in Renewable Energy Systems

This paper presents an intelligent computing model for wind speed prediction, which uses back propagation algorithm. Wind energy is inexhaustible unlimited clean energy. Wind power in the world has been rapidly growing. The prediction of wind speed has an important role in wind energy. The back-propagation algorithm (BPA) is used in the majority of neural networks application. The objective of this paper is to compute predicted output (wind speed) based on BP algorithm. The results are obtained using back propagation algorithm by training and testing methodologies. Simulation results show the performance of ANN for predicting wind speed in renewable energy systems

Prediction of Fracture Force of Weld Line Based on Artificial Neural Networks

A feed-forward three-layer neural network was proposed to predict the fracture force of injection-molded parts’ weld line. Firstly, the most significant process parameters which affect the fracture force of weld line were analyzed. Secondly, melt temperature, injection pressure, holding pressure and holding time were chosen as import variables and the fracture force of weld line was chosen as output variable to construct artificial neural networks. Furthermore, the performance of ANN was evaluated and tested by its application to verification tests with process parameters randomly selected which all of them were not used in the network training. Results showed that the ANN predictions yield mean absolute percentage error (MAPE) in the range of 0.86％,and maximum relative error (MRE) in the range of 1.84％ for the test data set, and which can comparatively accurately reflect the influence relation of the injection process parameters on part’s quality index under the circumstance of data deficiencies.

Prediction of Quality Index of Injection-Molded Parts by Using Artificial Neural Networks

Based on the deep analysis of injection molding process and factors influenced on the quality of injection-molded parts, and considering of the non-linear, multi-variable coupling of process and the problem of data deficiencies, a feed-forward three-layer neural network was presented to predict the part weight based on the melt temperature, holding pressure and holding time, which is the most important quality index of injection-molded parts. Furthermore, an architecture optimization approach for neural networks was used and the performance of ANN was evaluated and tested by its application to verification tests with process parameters randomly selected which all of them were not used in the network training. Results showed that the ANN predictions yield mean absolute percentage error (MAPE) in the range of 0.1％ and the maximum relative error (MRE) in the range of 0.47％ for the test data set, and which can comparatively accurately reflect the influence relation of the injection process parameters on part’s quality indexs under the circumstance of data deficiencies.

Modeling of neural image compression using GA and BP: a comparative approach

It is well known that the classic image compression techniques such as JPEG and MPEG have serious limitations at high compression rate; the decompressed image gets really fuzzy or indistinguishable. To overcome problems associated with conventional methods, artificial neural networks based method can be used. Genetic algorithm is a very powerful method for solving real life problems and this has been proven by applying to number of different applications. There is lots of interest to involve the GA with ANN for various reasons at various levels. Trapping in the local minima is one of the well-known problems of gradient decent based learning in ANN. The problem can be addressed using GA algorithm. But no work has been done to evaluate the performance of both learning methods from the image compression point of view. In this paper, we investigate the performance of ANN with GA in the application of image compression for obtaining optimal set of weights. Direct method of compression has been applied with neural network to get the additive advantage for security of compressed data. The experiments reveal that the standard BP with proper parameters provide good generalize capability for compression and is much faster compared to earlier work in the literature, based on cumulative distribution function. Further, the results obtained shows that general concept about GA, it performs better over gradient decent based learning, is not applicable for image compression.

Comparing Accuracy Performance of ANN, MLR, and GARCH Model in Predicting Time Deposit Return of Islamic Bank

Comparing Accuracy Performance of ANN, MLR, and GARCH Model in Predicting Time Deposit Return of Islamic Bank

Performance comparison of artificial neural networks and expert systems applied to flow pattern identification in vertical ascendant gas–liquid flows

Instantaneous readouts of an electrical resistivity probe are taken in an upward vertical air–water mixture. The signals are further processed to render the statistical moments and the probability density functions here used as objective flow pattern indicators. A series of 73 experimental runs have its flow pattern identified by visual inspection assisted by the analyses of the void fraction’s trace and associated probability density function. The flow patterns are classified into six groups and labeled as: bubbly, spherical cap, slug, unstable slug, semi-annular and annular. This work compares and analyzes the performance of artificial neural networks, ANN, and expert systems to flow pattern identification. The employed ANNs are Multiple Layer Perceptrons, Radial Basis Functions and Probabilistic Neural Network, with single and multiple outputs. The performance is gauged by the percentage of right identifications based on experimental observation. The analysis is extended to clustering algorithms to assist the formation of knowledge base employed during the learning stages of the ANNs and expert systems. The performance of the following clustering algorithms: self organized maps, K-means and Fuzzy C-means are also tested against experimental data.

Impact of the length of observed records on the performance of ANN and of conceptual parsimonious rainfall-runoff forecasting models

Although attractive to hydrologists, artificial neural network modeling still lacks norms that would help modelers to create and train efficient rainfall-runoff models in a systematic way. This study focuses on the impact of the length of observed records on the performance of multiple-layer perceptrons (MLPs), and compare their results with those of a parsimonious conceptual model equipped with an updating scheme. Both models were assessed for 1-day-ahead stream flow predictions. Ninety-two different model scenarios were obtained for 1-, 3-, 5-, 9-, and 15-year time sub-series created from a 24-year training set, shifting by a 1-year sliding window. All the model scenarios were verified against the same 7-year test set. The results revealed that MLP stream flow mapping was efficient as long as wet weather data were available for the training; the longer series implicitly guarantee that the data contain valuable information of the hydrological behavior; the results were consistent with those reported for conceptual rainfall-runoff models. The physical knowledge in the conceptual models allowed them to make much better use of 1-year training sets than the MLPs. However, longer training sets were more beneficial to the MLPs than to the conceptual model. Both types shared best performance about evenly for 3- and 5-year training sets, but MLPs did better whenever the training set was dominated by wet weather. The MLPs continued to improve for input vectors of 9 years and more, which was not the case of the conceptual model.

3720 ARTIFICIAL NEURAL NETWORK BASED PREDICTION OF OUTCOME OF LOWER GASTROINTESTINAL BLEEDING - A POTENTIAL TOOL FOR TRIAGE.

ANN based predictive models have been found to be useful for predicting outcome of different diseases.We evaluated if ANN can predict clinical outcome in patients with LGB based on information typically available during triage. Methods: Data on all patients admitted for LGB during 1998 were prospectively collected. A multi-layered perceptron neural network was constructed using clinical input variables which were available to the clinician during initial evaluation. The ANN was trained by back propagation using data from patients admitted during first half of 1998 (Group I) and then validated using data from patients admitted during the second half of 1998 (Group II). A stepwise logistic regression model was also constructed using available clinical input variables. Moreover, a previously validated scoring system (BLEED score, Crit Care Med 1997) was used to classify patients into high and low risk groups. Rebleeding, death during the same hospital admission and endoscopic, radiological or surgical intervention for control of hemorrhage were the outcome variables used to compare the different predictive models. Results: 141 patients (Mean age 74.9 years, 47 males) were available for analysis. Demographic and clinical features in Group I (n= 75) and Group II (n = 66) were similar. Diverticular bleeding was the predominant cause of bleeding (88, 62%). 7 (4.9%) patients expired, 26 (18.4%) had rebleeding and 21 (14.9%) needed intervention for control of bleeding. Performance of ANN in predicting the outcome variable were significantly better than the BLEED score (McNemar s test, P <0.01) and was comparable to that of the logistic regression based model (Table). Area under the receiver operating curve for ANN were 0.83, 0.87 and 0.91 respectively for rebleeding, intervention and death. Conclusion: ANN can predict outcome of patients with LGB reasonably well and may be used for risk stratification from information available at the time of initial evaluation.

PERFORMANCE EVALUATION OF BACKPROPAGATION NEURAL NETWORK IMPLEMENTATION ON SINGLE PROGRAM MULTIPLE DATA (SPMD) ARCHITECTURE

The Single Program Multiple Data (SPMD) approach has been developed to reduce the system overheads, to entail minimal computation time and to reduce the programming complexity of parallel programs. The SPMD computational model is based on the premise that all the processes execute the same program; however, at any instant of time, these processes may be executing different instructions and may be operating on different data. This work uses the SPMD computational model to implement a backpropagation neural network and the implementation is done on a multiprocessor. We present the distributed backpropagation forward and backward execution algorithms in SPMD mode. The performance of the SPMD-implemented ANN is demonstrated by the Automatic Target Recognition (ATR) problem, which involves identification of a class of targets depending on a set of input data values. The performance of the SPMD-implemented ANN is studied by observing the speedups achieved for training times as the number of processors is increa...