Search

This paper presents the application of an artificial neural network (ANN) to develop a model able to predict the holdup of a two-phase flow composed of water and mineral oil in a horizontal pipe. The surface velocities of each fluid and the pressure differential in the pipe are used as input parameters of the multi-layer artificial neural network with backpropagation, while the volumetric fraction of the fluids is used as an output parameter in the training. The 56 experimental data were obtained in the laboratory LabPetro-CEPETRO-UNICAMP. The results of the predictive model with the best performance show a mean absolute error (AAPE) of 3.01 % and a coefficient of determination R² of 0.9964, using 15 neurons in the hidden layer of the network and the TanSig transfer function. [ABSTRACT FROM AUTHOR]

This paper presents comparison of Standardization and Scaling methods in cost predicting. Four methods were used for pre-processing dataset; after that, data was processed through artificial neural network (RNA). The first step was to build common variables in information projects based on opinions of some project managers. Second step was to simulate dataset based on information provided by CRConsulting. Third one was process data with machine learning according to the four methods proposed, RNA algorithms were the same in four cases. Last, the comparison results were presented through adjustment models according to the applied method. [ABSTRACT FROM AUTHOR]

University dropout is a problem related to the student, as a direct responsible, and with the university institution, knowing the possibilities of attrition is relevant for the institution. In this paper, it is proposed to use classification models to find patterns and predict possible dropouts in university students. An application has been implemented that uses information provided by the university and that generates classification models from different algorithms (neural networks, ID3, C4.5), and uses the most significant attributes within the available information. The performance of these models was compared to define the one that provided the best results and that will be used to classify the students. The results show that the algorithm of C4.5 presented improvements in performance with respect to the neural network and the ID3 and that the relation of credits approved by a student related to the credits that he should have taken is the most significant variable in the construction of the model, followed by the qualifications, while the modality of the admission exam through which the student entered the university did not turn out to be significant, since it does not appear in the generated decision tree. [ABSTRACT FROM AUTHOR]

This paper addresses the «Regionalization of Perinatal Attention» in the Greater Buenos Aires. It analyses the state capabilities of Health Regions (HR) facing the challenge of coordinatingwork within regional health care/service networks, with emphasis in the HR 7. The methodological strategy is qualitative. The HR shows weaknesses in terms of formal institutions and resources, as well as by observing aspects related to the design of the policy and the relationship with the national level. Among the advances, this article shows that there are shared guiding ideas among the actors about what is the problem they need to address and how to do it, and that a certain network conscience exists. It emphasizes the importance of the individual abilities of managers to establish informal links and solutions case by case. [ABSTRACT FROM AUTHOR]

This paper is an advance of the research project entitled "Development of hybrid algorithms for data mining" and presents the use of a neural network with the simulated annealing algorithm to perform the prediction of a training data set. First, it addresses the problem to be solved, which is oriented to the analysis of the techniques defined for the hybrid algorithm. Then, the applied research methodology (descriptive-exploratory scientific with experimental approach) is justified. We performed a review of the techniques selected for the hybrid neuronal networks and simulated annealing technique which is applied to a set of experimental data associated with determining in a group of patients whether their spine is normal or abnormal. Then, the tests of analysis and results are presented. [ABSTRACT FROM AUTHOR]

This paper presents a controller performance that is develop employing an adaptive B-spline neural network algorithm for adjusting the rotor speed of the permanent magnet synchronous motor. It includes a comparative analysis with three control strategies: conventional proportional integral, sliding mode and fuzzy logic. Also, gives a systematic way to determine the optimal control gains and improve the tracking error performance. A methodology for the adaptive controller and its training procedure are explained. The efficacy of the proposed method is analyzed using time simulations where the motor is subjected to disturbances and reference changes. The proposed control technique exhibits the best performance because it can adapt to every condition, demanding low computational effort for an on-line operation and considering the system nonlinearities. [ABSTRACT FROM AUTHOR]

Optimal operation of energetic installations must considerably influence on its efficiency. By this, the objective of the present paper consists in the definition of the required models structure for the operation of downdraft gasifiers and its identification based on systemic analysis, it identification and the determination of the better adjustment type of model. For the models identification an experimental installation was constructed using, between other biomasses, firewood. Starting from an 3N experimental plan, with three replicas, needed models were identified helped by artificial neural networks and regression techniques. The best results were obtained using neural network, being therefore the most advisable in the preliminary stage of the process of the adaptive construction of operation models of the studied process. [ABSTRACT FROM AUTHOR]

In this paper we address the problem of automatic emotion recognition and classification through video. Nowadays there are excellent results focused on lab-made datasets, with posed facial expressions. On the other hand there is room for a lot of improvement in the case of 'in the wild' datasets, where light, face angle to the camera, etc. are taken into account. In these cases it could be very harmful to work with a small dataset. Currently, there are not big enough datasets of adequately labeled faces for the task. We use Generative Adversarial Networks in order to train models in a semi-supervised fashion, generating realistic face images in the process, allowing the exploitation of a big cumulus of unlabeled face images. [ABSTRACT FROM AUTHOR]

Environmental noise is a big problem related to the environmental pollution in cities, which affects the quality of people life. In this paper, a methodology that uses an approach based on Volunteered Geographic Information (VGI) for the monitoring, analysis and prediction of environmental noise is proposed. It can be very useful to propose alternatives and initiatives that improve the life in a city. So, this work is composed of the following stages: data acquisition, analysis and, data processing, as well as the information visualization, considering the temporality of the same and taking into account macro and micro levels of analysis for the study surface. In addition, some details of the design and development of a geographic information system are presented, consisting of a web-mapping system, an application for mobile devices called "NoiseMonitor", geospatial analysis and machine learning methods (support vector machines and artificial neural networks) for the prediction of environmental noise; by using contextual information; that is, some data related to the city. This kind of work seeks to take advantage of the willingness of citizens to participate collaboratively to sense their environment and be considered as human sensors, which unlike traditional approaches, the cost associated with the development and implementation of this project is much lower. Likewise, a case study based on the Mexico City is presented and discussed, particularly the fourth quadrant of the Historic Center of the City, which is very representative for the variety of environmental noise that is generated in that area. The application domain of this approach is oriented towards big data from a collaborative perspective, Internet of Things and smart cities. [ABSTRACT FROM AUTHOR]

Los sensores de reflexión ópticos para medir distancia se caracterizan por la no linealidad de la salida. Esta condición deben asumirla las unidades de procesamiento donde son utilizados. El presente artículo tiene por objeto reducir el costo computacional, en términos de espacio de memoria y tiempo de procesamiento, en la linealización de la salida de los sensores para ser usados en sistemas embebidos de bajo costo. Los resultados de dos diferentes sensores indican que la curva exponencial estimada se puede ajustar usando redes neuronales con funciones de base radial con una reducción de más del 50 % en el tiempo de procesamiento. Para el diseño de la red se estudió la eficiencia respecto al tamaño de la red y a la distribución de los centroides con resultados importantes. La reducción del tiempo y de espacio lograda permite aprovechar los recursos del sistema en otras tareas, al mismo tiempo que posibilita aumentar la frecuencia de muestreo en la adquisición. [ABSTRACT FROM AUTHOR]

This paper offers a comparative analysis of the effectiveness of eight popular forecasting methods: univariate, linear, discriminate and logit regression; recursivepartitioning, rough sets, artificial neural networks, and DEA. Our goals are: clarify the complexity-effectiveness balance of each methodology; identify a reduced set of independent variables that are significant predictors whatever the methodology is; and discuss and relate these findings to the financial theory, to help consolidate the foundations of a theory of financial failure. Our results indicate that, whatever the methodology is, reliable predictions can be made using four variables; these ratios convey informatión about profitability, financial structure, rotation, and operating cash flows. [ABSTRACT FROM AUTHOR]

In this paper is shown the application of neural networks in order to predict academic marks that will be obtained for the students in the subjects of Data Structures I and II, included both in the Informatics Engineering curricula at Higher Polytechnic Institute José Antonio Echeverría in the Republic of Cuba. The main motivation for the present work is justified because selected subjects have a high level of complexity, demanding from the student to be rigorous and a permanent dedication. As a consequence the academic results obtained at the present time are not satisfactory. To reach the goal mentioned above a software based on MATLAB tool was developed and the marks obtained previously by students in some subjects and others data of interest were used. Two neural networks were employed, both with the same architecture, but each one trained with the specific data of each subject (Data Structures I and II). A group of experiments was carried out to contrast the behavior of the neural networks regarding some specific statistics in the data of the sample. An overall effectiveness in prediction superior to 78% for the case of the first subject was achieved, while for the second one effectiveness superior to 75% was reached. [ABSTRACT FROM AUTHOR]

This paper shows the development and use of a hybrid algorithm of machine learning for the missing values predict task, with this task that is carried out while the data preprocessing phase. Firstly, we go through the problem to solve, which is pointed to the study and analysis of different techniques for the missing values predict in order to suggest a hybrid technique as a product of this research for the task and link it to the OLAP (On-Line Analytical Processing) technology. Then, justifying the research methodology (scientific descriptive - probing) applied in this project. The techniques review was carried out filling out the missing values; based on the techniques proof and the study cases, k-Nearest Neighbors algorithms and artificial neural networks were selected and a hybrid technique (KMediaSom) was suggested, applied to a synthetic data set and a real one, coming from a OLAP; for the algorithms implementation was used Matlab. Right away, the analysis and results are set out in order to specify its applicability about efficacy and time complexity. Results are suitable as for the synthetic data set as for the real one; according to the test signs achieved. Finally, the conclusions, where it i's proved that the technique or hybrid algorithms generate better results than the techniques used by separately. [ABSTRACT FROM AUTHOR]

Time series prediction is an important research problem due to its implications in engineering, economics, finance and social sciences. An important topic about this problematic is the development of new models and its comparison with previous approaches in terms of forecast accuracy. Recently, support vector machines (SVM) have been used for time series prediction, but the reported experiences are limited and there are some problems related to its specification. The aim of this paper is to propose a novel technique for estimating some constants of the SVM usually fixed empirically by the modeler. The proposed technique is used to estimate several SVM with the aim of forecast five benchmark time series; the obtained results are compared with the statistics reported in other papers. The proposed method allow us to obtain competitive SVM for the time series forecasted in comparison with the results obtained using other most traditional models. [ABSTRACT FROM AUTHOR]

Cancer is a health problem, resulting in the first cause of death worldwide. In the present paper, a Quantitative Structure-Activity Relationship (QSAR) study was developed as a tool for the design of new drugs with potential antitumor activity. Discriminant Linear Analysis and Neural Network were the mathematics methods used to estimate the activity of in a data set consisting in 300 compounds. The biological activity, extracted from the US National Cancer Institute was divided by cluster analysis in a training and prediction series. A model with 10 variables and 84,33% of correct classification was obtained by a discriminant function meanwhile, the neural network tested with the same number of variables resulted in a 89,67% of accuracy. Also was calculated the contribution of different structural fragments on the cytostatic activity, and quantified their contribution. Six new compounds were designed predicting a good antitumor activity. In general, the predictive quality of the neural network model was higher than the linear discriminant. [ABSTRACT FROM AUTHOR]

Biomass, as an alternative source of energy, can be converted into biogas through the anaerobic digestion of organic material. The device used for this purpose is called biodigestor, and several protocols have been studied aimed at estimating the volume and quantity of the gas obtained. These protocols use a combination of the origin of the biomass and equations involving environment temperature, biological load and retention time. The treatment of organic residues and the systematization of the activity on livestock farms requires the intervention of expert users or the implementation of intelligent systems, which allow for quantitative and qualitative variables. In this paper, we use Artificial Neural Networks (ANN) to estimate the production volume of a bio-digester ball. The networks are designed for this use on both livestock and farming residues. The efficiency of the estimation is evaluated using R Correlation Coefficient. Results show that the technique is reliable and can be used in practice. [ABSTRACT FROM AUTHOR]

A system that combines the use of Wavalet Transforms (WT), Linear Prediction Coefficients (LPC) and Artificial Neural Networks (ANN) are shown in this paper. Vowels and syllables recognition speaker independent can be made. Using this structure, automated software is proposed, which through an interface, allows users with hearing difficulties or total absence of this, the possibility of using it primarily as an initial tool to support learning of syllables. In the first stage, a small number of syllables is incorporated, especially those that may have more difficulty in their identification. Subsequently, it can be incorporated a larger number of syllables, then it can be growing. Adding new syllables, would allow (through segmenting words into syllables) a greater deployment system, for identifying complete words and therefore the spoken language learning. [ABSTRACT FROM AUTHOR]

This paper presents the development of a system aimed to facilitate the communication and interaction of people with severe hearing impairment with other people. The system employs artificial vision techniques to the recognition of static signs of Colombian Sign Language (LSC). The system has four stages: Image capture, preprocessing, feature extraction and recognition. The image is captured by a digital camera TRDB-D5M for Altera's DE1 and DE2 development boards. In the preprocessing stage, the sign is extracted from the background of the image using the thresholding segmentation method; then, the segmented image is filtered using a morphological operation to remove the noise. The feature extraction stage is based on the creation of two vectors to characterize the shape of the hand used to make the sign. The recognition stage is made up a multilayer perceptron neural network (MLP), which functions as a classifier. The system was implemented in the Altera's Cyclone II FPGA EP2C70F896C6 device and does not require the use of gloves or visual markers for its proper operation. The results show that the system is able to recognize all the 23 signs of the LSC with a recognition rate of 98.15 %. [ABSTRACT FROM AUTHOR]

This paper offers an exhaustive analysis of the effectiveness of several models and methodologies that are commonly used to forecast financial failure: Linear, M DA, Logit, and artificial neural network. Our main aim is to evaluate their relative strengths and weaknesses, in terms of technical reliability and error cost; to do so, models are estimated and validated, and then used to perform an artificial simulation to evaluate which of them causes the lower cost of errors. Reliability is examined in four forecast horizons, to collect evidences about temporal (in) stability. We also check the relative advantages of financial ratios-based models, versus audit-based forecast models. Our results suggest that all models attain a high performance rate; however, artificial neural networks'forecasts seem to be more stable, both in temporal and cross-sectional perspectives. [ABSTRACT FROM AUTHOR]

Electric arc furnace provides a relatively simple way for melting metals. They are used in the production of highly purified steel, aluminium, copper and other metals. However, they are considered the more damaging load for the power system. It is very important, therefore, to count on arc furnace models for determining with high degree of accuracy the performance of this type of load. In this way, it would be possible to assess the impact in terms of power quality indices for the power system to which they might be connected. When using electric arc furnace models in practice, a key issue is the calibration of the parameters of the model. In this paper, we show a procedure for calibrating all the parameters of an AC electric arc furnace model using real measurements of voltages and currents. It uses a multilayer neural network as an emulator of the electric arc furnace model. The neural network is trained using data obtained from the simulation of the electric arc furnace model implemented in Matlab®-Simulink®. Once the network is trained, the parameters of interest are obtained by solving an inverse problem. Results obtained show a maximum percentage error of 4.1 % for the rms value of the current involved in the electrical arc. [ABSTRACT FROM AUTHOR]

Optical recognition of handwritten numbers was worked by different methods, with satisfactory results. In this paper, we propose fuzzy systems with genetic algorithms, specifically memetic, to do this task. Results with this method are compared with neuronal network of semi-supervised learning, where we was used networks with radial base functions (RBF). To make the comparison, is possible observed that this kind of neuronal network offers advantages regarding error rates and time-to-results of the recognition system, compared with methods based in fuzzy systems. [ABSTRACT FROM AUTHOR]

This paper presents a strategy for a generator to participate and to mitigate the risk effect of price volatility in the Colombian wholesale electricity market. The strategy is used to optimize the generator participation in the long-term market (bilateral market) and the spot market. Additionally, the strategy mitigates the risk of price exposure in the sport market using electricity forward contracts. Numerical results shows that the proposed methodology is more efficient than classical optimization models since this proposal considers the intrinsic price volatility of the long-term and spot markets. [ABSTRACT FROM AUTHOR]

One of the most important uses of artificial neural networks is to forecast non-linear time series, although model-building issues, such as input selection, model complexity and parameters estimation, remain without a satisfactory solution. More of research efforts are devoted to solve these issues. However, other models emerged from statistics would be more appropriated than neural networks for forecasting, in the sense that the process of model specification is based entirely on statistical criteria. Multivariate adaptive regression splines (MARS) is a statistical model commonly used for solving nonlinear regression problems, and it is possible to use it for forecasting time series. Nonetheless, there is a lack of studies comparing the results obtained using MARS and neural network models, with the aim of determinate which model is better. In this paper, we forecast four nonlinear time series using MARS and we compare the obtained results against the reported results in the technical literature when artificial neural networks and the ARIMA approach are used. The main finding in this research, it is that for all considered cases, the forecasts obtained with MARS are lower in accuracy in relation to the other approaches. [ABSTRACT FROM AUTHOR]

This paper is related to the application of Artificial Neural Networks (ANN) for noise cancelation in electric digital signals. The use of ANN is a novel technique in noise reduction, which only needs the noisy version of the signal to obtain the clean one. In this work is employed a FIR Multilayer Perceptron network. First is applied a software approach in Matlab, where the network is trained, and it's designed a strategy to reach the optimal network for noise cancelation in two cases study: simulated noise in Matlab and noise obtained through a Data Acquisition System connected to a sensor. Good results were achieved in the cancelation of both noises. Second approach is used to describe the ANN in hardware. The optimal FIR Multilayer Perceptron architecture of Matlab is implemented in VHDL (Very High Speed Integrated Circuit Hardware Description Language) to download in a Xilinx XC3S1200E FPGA (Field Programmable Gate Array), setting as goal in the design highlight the parallelism of ANN operation. After been obtained the VHDL design, a noise cancelation application is simulated, with good results, considering the errors produced by the less accuracy of the numerical format of VHDL ANN. Finally the design is downloaded in the FPGA and is checked that works according to the results of the software approach. [ABSTRACT FROM AUTHOR]

The study of seismic intensity attenuation plays an important role in the analysis of menace that includes historical events. Mapping the intensity attenuation is usually done by regressing the intensity versus distance. Nowadays there are different ways to examine the characteristics of a seismic event using instrumental means, however, the experts face the problem of qualitative nature of the sources of information and the mapping of the relationship between intensity, magnitude and distance for the generation of risk scenarios based on historical information. This paper presents an alternative to map this relationship through artificial neural networks (ANN). As a result, we propose a procedure that was validated through the mapping of the intensities of 68 earthquakes that occurred northern South America, between 1766 and 2004. We found that ANNs present advantages with respect to the conventional models of regression: a. they preserve in a better way the first statistical moment, b. they reflect a minor approximation error and c. the variance explained by the ANN is better than the one from the models of statistical regression. [ABSTRACT FROM AUTHOR]

In this paper we propose a new methodology for the prediction of nonlinear time series using genetic programming. The proposed approach is based on incorporating the concept of functional blocks and the modification of the genetic algorithm so that it operates with it. The functional blocks represent well known statistical models for the time series forecasting. The proposed algorithm allows the exploration and exploitation of regions where there is greater possibility of finding better forecasting models. Two Benchmark time series were predicted in order to validate the proposed approach, and it was found that our methodology predicts more accurately the time series considered, in comparison with other nonlinear models. [ABSTRACT FROM AUTHOR]

In this paper one method for bearings diagnosis is proposed and evaluated. This method use signal pattern recognition from mechanical vibrations. Wavelet and Fourier transforms are used for pre-processing the signal and an Artificial Neural Network (ANN) is used as a classifier. Analysis of variance (ANOVA) is used for evaluating the ANN inputs. ANOVA is performed to compare the effect of the factors: speed, load, outer race fault and rolling element fault on each of the parameters proposed as inputs of the ANN, looking for the best parameters for classifying the faults. About 2000 ANN structures were trained in order to find the most appropriate classifier. The results show that the average of success in classifying was 88,5 % for the scaled conjugate gradient algorithm (trainscg), while the Levenberg Marquardt algorithm (trainlm) presented 91,8 %. Besides, it was possible to achieve 100 % of success in classifying in 7 cases. [ABSTRACT FROM AUTHOR]

This paper presents an Artificial Neural Network (ANN) based model for environmental variables related to the tobacco drying process. A fitting ANN was used to estimate and predict temperature and relative humidity inside the tobacco dryer: the estimation consists of calculating the value of these variables in different locations of the dryer and the prediction consists of forecasting the value of these variables with different time horizons. The proposed model has been validated with temperature and relative humidity data obtained from a real tobacco dryer using a Wireless Sensor Network (WSN). On the one hand, an error under 2% was achieved, obtaining temperature as a function of temperature and relative humidity in other locations in the estimation task. Besides, an error around 1.5 times lower than the one obtained with an interpolation method was achieved in the prediction task when the temperature inside the tobacco mass was predicted with time horizons over 2.5 hours as a function of its present and past values. These results show that ANN-based models can be used to improve the tobacco drying process because with these types of models the value of environmental variables can be predicted in the near future and can be estimated in other locations with low errors. [ABSTRACT FROM AUTHOR]

Companies engaged in the manufacture, marketing and maintenance of cylinders for liquefied petroleum gas in Colombia, stamped steel plates and welded to the product a unique serial code to be identified within the cylinder of the country park. Currently, the identification process is manual and checked approximately 7000 cylinders per day in a single factory. The main objective of this paper is to present a vision system that uses artificial neural networks to recognize the code. This system consists physically of a portable device that controls light environment and scene for the acquisition of images. Another component of the system is to adjust the image. The adjustment is based on median filtering, binarization, label, and segmentation, this processing allows more meaningful information and image discrimination. Finally, the intelligent component identification is performed with Hopfield neural networks and an algorithm that checks the development of image recognition. The effectiveness of the system was reported with experimental results obtained on the basis of error with a significant number of samples. [ABSTRACT FROM AUTHOR]

Clays are the main raw material in the manufacture of products for the construction sector, such as tile, veneer, flooring and bricks. Small and medium enterprises generally use brick clays of different mineralogical origin, classified in order to formulate their mixtures according to the production team experience; the uncertainty associated with this method causes that a portion of their manufactured products are rejected, because their properties do not meet the technical specifications. This paper presents a methodology based on neural networks for classification of clays, based on the clay properties to be used to make the pasta, with the aim of reducing the number of rejected products. It used different network topologies for classification, and chose the one, which have been found capable to predict the training and testing samples with an accuracy of 97.79 % and 94.12 %, respectively. [ABSTRACT FROM AUTHOR]

Many time series with trend and seasonal pattern are successfully modeled and forecasted by the airline model of Box and Jenkins; however, this model neglects the presence of nonlinearity on data. In this paper, we propose a new nonlinear version of the airline model; for this, we replace the moving average linear component by a multilayer perceptron neural network. The proposed model is used for forecasting two benchmark time series; we found that the proposed model is able to forecast the time series with more accuracy that other traditional approaches. [ABSTRACT FROM AUTHOR]

In this paper is presented a multilayer perceptron neural network combined with the Nelder-Mead Simplex method to detect damage in multiple support beams. The input parameters are based on natural frequencies and modal flexibility. It was considered that only a number of modes were available and that only vertical degrees of freedom were measured. The reliability of the proposed methodology is assessed from the generation of random damages scenarios and the definition of three types of errors, which can be found during the damage identification process. Results show that the methodology can reliably determine the damage scenarios. However, its application to large beams may be limited by the high computational cost of training the neural network. [ABSTRACT FROM AUTHOR]

In this paper we develop a back-propagation learning algorithm for feedforward neural networks trained online. Three neurocontrollers are designed for three systems. Those systems are an RC circuit, a DC motor (electronically emulated) and a sphere-tube system. The first implemented strategy is a standard PID controller, which is used in order to compare the performance of the neurocontrollers. The first neurocontroller leads the system in parallel with a PID; the next one is trained online to work alone, and the last one is a neural PID, which strives to make the controller adaptable to the dynamics of the plant trough changes on the PID gains. The control is carried out in real time by using Simulink and a PCI 6024E data acquisition card. The results for each system are also included. [ABSTRACT FROM AUTHOR]

This paper presents an application of the PLC in the implementation of advanced control systems to regulate a plant of level. The process is identified and the work is based on three control algorithms: a control with neural networks, a RST controller and a PI controller estimated using pole placement. It is developed a program for the PLC which, together with the SCADA system, allows using these techniques of control in the regulation of industrial processes. The Software also includes the design of a user-friendly environment for viewing the variables from the user. The results obtained in the control of the process, demonstrate the likelihood to implement advanced control algorithms with very good performance in the PLC. [ABSTRACT FROM AUTHOR]

The motion estimation is an essential block in the prediction stage of the H.264/AVC video compression standard, which is crucial for obtaining an effective coding rate. Different algorithms have been proposed in the literature to optimize the architecture of this block, since it requires a long execution time and a significant number of computational resources. Traditional strategies based on the search on each block allow ensuring high quality but are inappropriate in terms of computational efficiency; other methods based on search by regions reduce the processing load, but do not guarantee a good quality. In this paper it is implemented a multilayer perceptron neural network with supervised backpropagation learning algorithm which allows making the estimation. [ABSTRACT FROM AUTHOR]

Every day organizations have more information because their systems produce a large amount of daily operations which are stored in transactional databases. In order to analyze this historical information, an interesting alternative is to implement a Data Warehouse. In the other hand, Data Warehouses are not able to perform predictive analysis for themselves, but machine learning techniques can be used to classify, grouping and predict historical information in order to improve the quality of analysis. This paper depicts architecture of a Data Warehouse useful to perform an analysis of students' academic performance. The Data Warehouse is used as input of a Neural Network in order to analyze historical information and forecast. The results show the viability of using Data Warehouse for academic performance analysis and the feasibility of predicting the number of approved courses for students using only their own historical information. [ABSTRACT FROM AUTHOR]

This article shows that artificial neural networks are used for confirming the relationships between physiological and cognitive changes. Specifically, we explore the influence of a decrease of neurotransmitters on the behaviour of old people in recognition tasks. This artificial neural network recognizes learned patterns. When we change the threshold of activation in some units, the artificial neural network simulates the experimental results of old people in recognition tasks. However, the main contributions of this paper are the design of an artificial neural network and its operation inspired by the nervous system and the way the inputs are coded and the process of orthogonalization of patterns. [ABSTRACT FROM AUTHOR]

This paper discusses how two artificial intelligence techniques were combined, neural networks and genetic algorithms for the development of a computational tool used for the estimation of mechanical properties such as tensile strength, uniaxial compressive strength and triaxial compressive strength in sandstones, from petrophysical properties using data from tests of Rock Mechanics Laboratory of the Colombian Petroleum Institute - Ecopetrol SA as training data, to improve the design of non-destructive testing with some degree of confidence and resulting in cost reduction. [ABSTRACT FROM AUTHOR]

Artificial neural networks are an important technique in nonlinear time series forecasting. However, training of neural networks is a difficult task, because of the presence of many local optimal points and the irregularity of the error surface. In this context, it is very easy to obtain under-fitted or over-fitted forecasting models without forecasting power. Thus, researchers and practitioner need to have criteria for detecting this class of problems. In this paper, we demonstrate that the use of well known methodologies in linear time series forecasting, such as the Box-Jenkins methodology or exponential smoothing models, are valuable tools for detecting bad specified neural network models. [ABSTRACT FROM AUTHOR]

In order to capture the maximum energy from the wind, control systems operating always at an optimum power has been utilized. This control system, considered as a whole, must be able of respond with an adequate precision and speed in response to the randomness and variability of the wind. The relationship between the wind speed, the blade pitch and the generator speed in order to produce the maximum power and also be able to limit the output power for large wind speeds is a very complicated one and it is very difficult to find its mathematical function. In this paper, the authors, utilizing the MATLAB SIMULINK toolboxes, propose representing this functional relation by means of an Artificial Neural Network (ANN). The parameters and characteristics of an existing wind turbine generator are utilized and it is demonstrated that it is possible to use an ANN in the simulation and control of a variable speed, variable pitch wind turbine that capture the maximum power from the wind. [ABSTRACT FROM AUTHOR]

This paper analyses the influence of some variables in the shelf space occupied by store brands. We propose and test a theoretical model of store brand shelf space. Data were collected for 29 product categories in 55 retail stores. A two-phase procedure was adopted: (1) multiple regression analyses; (2) neural network simulation (ANN). The application of this last method improves the goodness of fit obtained through the regression method. Furthermore, it presents additional advantages since ANN does not need to fulfil the main assumptions needed in regression analyses. The findings corroborate our proposed model, in that all hypothesized relationships and directions are supported. On this basis, we draw theoretical as well as useful managerial implications for both retailers and manufacturers. [ABSTRACT FROM AUTHOR]

This paper described the problem of path-planning in mobile robots in a partially structured environment. The work was divided in three parts: design of a mechatronic system, denominated Robot IO, study of advantages and disadvantages of three path-planning algorithms: neural nets, genetic algorithms and fuzzy logic, and implementation of algorithms that allow to identify the workspace where IO moves. [ABSTRACT FROM AUTHOR]

This paper proposes a methodology to build hybrid models to fault location in power distribution systems, this methodology will allow to the utilities improve their quality indexes corresponding energy service continuity (DES and FES indexes). The hybrid model in methodology is composed by a technique based on knowledge (LAMDA tecnique) and other based on model (Ratan Das algorithm). The LAMDA technique is a technique based on artificial intelligence inheriting characteristics of the fuzzy logic and neural networks. The Rantan Das algorithm, is fault location algorithm that estimate the fault location from voltage and current phasors in the moment of fault and moreover some electric parameters of the distribution system. The novel thing of the methodology is centred in which with the implementation of the hybrid model is improved the precision in the estimation of the fault location, due to that is reduced the multiple estimation of the fault location algorithm by the technique based on knowledge. Finally, are presented the results obtained in test realized with a distribution circuit of 24 kV and length 60 km, approximately. [ABSTRACT FROM AUTHOR]

Modeling and forecasting exchange rates is a big economic headache. This paper uses an artificial neuronal network model to represent the Colombian real exchange index dynamics because it describes the series dynamics better than a self-regressive linear model does, as may be appreciated in the result of the verosimilitud radius contrast. The model was accepted after applying a series of standard tests and contrasting the results against those obtained using a self-regressive linear model. The results indicated that the current value of a series solely depends on its previous value. [ABSTRACT FROM AUTHOR]

Abstract: This paper employs artificial neural network and decision tree to derive knowledge about the job attitudes of Generation Xers. The sample frame consisted of 1000 large Manufacturing Industries and 500 large Service Industries, randomly selected from the Common Wealth Magazine 1000 index of Taiwan Manufacturing Industries and Service Firms. Then, we exploited the ART2 neural model to take the collected data as inputs and form performance classes according to their similarities. Finally, the decision tree was employed to determine definitions for each class, resulting in 52 rules associated with certainty factors. The results could be used to develop an intelligent decision support system for the recruitment and management of Generation Xers. [Copyright &y& Elsevier]

This paper analyzes the capacity of the neural networks to forecast the sign of weekly variations of IPSA. Several architectures of neural networks were used over the time period between January 11th of 1999 to October 22th of 2001, being the Recursive Ward Network the one with the best performance, reaching an outsample predictive capacity of 72% and an outsample accumulate yield for the IPSA portfolio of a 24.42%. The Recurrent Recursive Jordan & Elman Network achieved a forecast ability of 64% and a return of 21.33%; while the AR(1,1) model obtained a return of 18.31% higher than the Ward Standard Network and Recursive MLP returns. Even though the first one had not statistical evidence of predictive capacity it would allow to conclude that a higher predictive capacity do not always implies higher yields. The Pesaran & Timmermann test (1992) gives evidence that the Recursive Ward Network and the Recurrent Recursive Jordan & Elman Network, in his recursive and standard version can forecast the index directional change for the chilean case. Also, this models could produce higher returns than AR(1,1) model. This result supports the conclusions of Leung, Daouk & Chert (2000) about the prediction of movements directions can give greater capital gains that the forecasting of clove values. [ABSTRACT FROM AUTHOR]