Previsão do índice bovespa por meio de redes neurais artificiais: uma análise comparada aos métodos tradicionais de séries de tempo

Forecast is the basis for making strategic, tactical and operational business decisions. In financial economics, several techniques have been used to predict the behavior of assets over the past decades.Thus, there are several methods to assist in the task of time series forecasting, however, conventional modeling techniques such as statistical models and those based on theoretical mathematical models have produced unsatisfactory predictions, increasing the number of studies in more advanced methods of prediction. Among these, the Artificial Neural Networks (ANN) are a relatively new and promising method for predicting business that shows a technique that has caused much interest in the financial environment and has been used successfully in a wide variety of financial modeling systems applications, in many cases proving its superiority over the statistical models ARIMA-GARCH. In this context, this study aimed to examine whether the ANNs are a more appropriate method for predicting the behavior of Indices in Capital Markets than the traditional methods of time series analysis. For this purpose we developed an quantitative study, from financial economic indices, and developed two models of RNA-type feedfoward supervised learning, whose structures consisted of 20 data in the input layer, 90 neurons in one hidden layer and one given as the output layer (Ibovespa). These models used backpropagation, an input activation function based on the tangent sigmoid and a linear output function. Since the aim of analyzing the adherence of the Method of Artificial Neural Networks to carry out predictions of the Ibovespa, we chose to perform this analysis by comparing results between this and Time Series Predictive Model GARCH, developing a GARCH model (1.1).Once applied both methods (ANN and GARCH) we conducted the results' analysis by comparing the results of the forecast with the historical data and by studying the forecast errors by the MSE, RMSE, MAE, Standard Deviation, the Theil's U and forecasting encompassing tests. It was found that the models developed by means of ANNs had lower MSE, RMSE and MAE than the GARCH (1,1) model and Theil U test indicated that the three models have smaller errors than those of a naïve forecast. Although the ANN based on returns have lower precision indicator values than those of ANN based on prices, the forecast encompassing test rejected the hypothesis that this model is better than that, indicating that the ANN models have a similar level of accuracy . It was concluded that for the data series studied the ANN models show a more appropriate Ibovespa forecasting than the traditional models of time series, represented by the GARCH model

Aplicação de indicadores webométricos nos programas de Pós-Graduação das engenharias recomendados pela CAPES

This study aims to identify, through the application of webometric indicators, which Post-Graduate Courses in Engineering recommended by the Coordination of Improvement of Higher Personnel Education (CAPES) in Brazil stand out in the web space, in relation to the communication process and dissemination of scientific information in the academic environment. For this, we analyzed the structures content of the sites, the use, through the conduct of investigations and searches, the quality of information available, as well as the structure of existent hypertexts in the sites of this universe of search. The tools and methodologies adopted for this study are: search engines (Google, Yahoo), Mapper software (Xenu Link Sleuth) and analysis software and visualization of networks (and Ucinet6 NetDraw). Webometric indicators are also used, such as size of the web sites, visibility, web impact factor, brightness and density of the network. These instruments provide a brief analysis and evaluation for this webometric study. Therefore, from the incursion of the literature used, it appears that there are many advantages of using this type of metric study in the so called Information Society. The obtained results could identify which postgraduate courses in engineering has a better availability of their information on the Web, as well to define which of these courses stands out in relation to the use of their information, which has been outstanding in respect to its impact factor and which offers a greater number of links that serve as a source of information for its users, contributing, in its turn, with the navigability of the same network. In summary, it is asserted that the webometric study presents promising results, which are able to achieve the proposed objectives, as well as identify the factors that contribute significantly to the good visualization of these sites in the network, thus helping the spread of information and scientific communication through the use of the Web.

Indicadores de cálcio e de voltagem codificados geneticamente na detecção de potenciais de ação e inputs sinápticos em cultura de neurônios hipocampais

Recently, genetically encoded optical indicators have emerged as noninvasive tools of high spatial and temporal resolution utilized to monitor the activity of individual neurons and specific neuronal populations. The increasing number of new optogenetic indicators, together with the absence of comparisons under identical conditions, has generated difficulty in choosing the most appropriate protein, depending on the experimental design. Therefore, the purpose of our study was to compare three recently developed reporter proteins: the calcium indicators GCaMP3 and R-GECO1, and the voltage indicator VSFP butterfly1.2. These probes were expressed in hippocampal neurons in culture, which were subjected to patchclamp recordings and optical imaging. The three groups (each one expressing a protein) exhibited similar values of membrane potential (in mV, GCaMP3: -56 ±8.0, R-GECO1: -57 ±2.5; VSFP: -60 ±3.9, p = 0.86); however, the group of neurons expressing VSFP showed a lower average of input resistance than the other groups (in Mohms, GCaMP3: 161 ±18.3; GECO1-R: 128 ±15.3; VSFP: 94 ±14.0, p = 0.02). Each neuron was submitted to current injections at different frequencies (10 Hz, 5 Hz, 3 Hz, 1.5 Hz, and 0.7 Hz) and their fluorescence responses were recorded in time. In our study, only 26.7% (4/15) of the neurons expressing VSFP showed detectable fluorescence signal in response to action potentials (APs). The average signal-to-noise ratio (SNR) obtained in response to five spikes (at 10 Hz) was small (1.3 ± 0.21), however the rapid kinetics of the VSFP allowed discrimination of APs as individual peaks, with detection of 53% of the evoked APs. Frequencies below 5 Hz and subthreshold signals were undetectable due to high noise. On the other hand, calcium indicators showed the greatest change in fluorescence following the same protocol (five APs at 10 Hz). Among the GCaMP3 expressing neurons, 80% (8/10) exhibited signal, with an average SNR value of 21 ±6.69 (soma), while for the R-GECO1 neurons, 50% (2/4) of the neurons had signal, with a mean SNR value of 52 ±19.7 (soma). For protocols at 10 Hz, 54% of the evoked APs were detected with GCaMP3 and 85% with R-GECO1. APs were detectable in all the analyzed frequencies and fluorescence signals were detected from subthreshold depolarizations as well. Because GCaMP3 is the most likely to yield fluorescence signal and with high SNR, some experiments were performed only with this probe. We demonstrate that GCaMP3 is effective in detecting synaptic inputs (involving Ca2+ influx), with high spatial and temporal resolution. Differences were also observed between the SNR values resulting from evoked APs, compared to spontaneous APs. In recordings of groups of cells, GCaMP3 showed clear discrimination between activated and silent cells, and reveals itself as a potential tool in studies of neuronal synchronization. Thus, our results indicate that the presently available calcium indicators allow detailed studies on neuronal communication, ranging from individual dendritic spines to the investigation of events of synchrony in neuronal networks genetically defined. In contrast, studies employing VSFPs represent a promising technology for monitoring neural activity and, although still to be improved, they may become more appropriate than calcium indicators, since neurons work on a time scale faster than events of calcium may foresee

Distribuição de derivados de petróleo por redes de polidutos: uma abordagem através de algoritmos evolucionários híbridos para um problema triobjetivo

An important problem faced by the oil industry is to distribute multiple oil products through pipelines. Distribution is done in a network composed of refineries (source nodes), storage parks (intermediate nodes), and terminals (demand nodes) interconnected by a set of pipelines transporting oil and derivatives between adjacent areas. Constraints related to storage limits, delivery time, sources availability, sending and receiving limits, among others, must be satisfied. Some researchers deal with this problem under a discrete viewpoint in which the flow in the network is seen as batches sending. Usually, there is no separation device between batches of different products and the losses due to interfaces may be significant. Minimizing delivery time is a typical objective adopted by engineers when scheduling products sending in pipeline networks. However, costs incurred due to losses in interfaces cannot be disregarded. The cost also depends on pumping expenses, which are mostly due to the electricity cost. Since industrial electricity tariff varies over the day, pumping at different time periods have different cost. This work presents an experimental investigation of computational methods designed to deal with the problem of distributing oil derivatives in networks considering three minimization objectives simultaneously: delivery time, losses due to interfaces and electricity cost. The problem is NP-hard and is addressed with hybrid evolutionary algorithms. Hybridizations are mainly focused on Transgenetic Algorithms and classical multi-objective evolutionary algorithm architectures such as MOEA/D, NSGA2 and SPEA2. Three architectures named MOTA/D, NSTA and SPETA are applied to the problem. An experimental study compares the algorithms on thirty test cases. To analyse the results obtained with the algorithms Pareto-compliant quality indicators are used and the significance of the results evaluated with non-parametric statistical tests.

Previsão do índice bovespa por meio de redes neurais artificiais: uma análise comparada aos métodos tradicionais de séries de tempo

Forecast is the basis for making strategic, tactical and operational business decisions. In financial economics, several techniques have been used to predict the behavior of assets over the past decades.Thus, there are several methods to assist in the task of time series forecasting, however, conventional modeling techniques such as statistical models and those based on theoretical mathematical models have produced unsatisfactory predictions, increasing the number of studies in more advanced methods of prediction. Among these, the Artificial Neural Networks (ANN) are a relatively new and promising method for predicting business that shows a technique that has caused much interest in the financial environment and has been used successfully in a wide variety of financial modeling systems applications, in many cases proving its superiority over the statistical models ARIMA-GARCH. In this context, this study aimed to examine whether the ANNs are a more appropriate method for predicting the behavior of Indices in Capital Markets than the traditional methods of time series analysis. For this purpose we developed an quantitative study, from financial economic indices, and developed two models of RNA-type feedfoward supervised learning, whose structures consisted of 20 data in the input layer, 90 neurons in one hidden layer and one given as the output layer (Ibovespa). These models used backpropagation, an input activation function based on the tangent sigmoid and a linear output function. Since the aim of analyzing the adherence of the Method of Artificial Neural Networks to carry out predictions of the Ibovespa, we chose to perform this analysis by comparing results between this and Time Series Predictive Model GARCH, developing a GARCH model (1.1).Once applied both methods (ANN and GARCH) we conducted the results' analysis by comparing the results of the forecast with the historical data and by studying the forecast errors by the MSE, RMSE, MAE, Standard Deviation, the Theil's U and forecasting encompassing tests. It was found that the models developed by means of ANNs had lower MSE, RMSE and MAE than the GARCH (1,1) model and Theil U test indicated that the three models have smaller errors than those of a naïve forecast. Although the ANN based on returns have lower precision indicator values than those of ANN based on prices, the forecast encompassing test rejected the hypothesis that this model is better than that, indicating that the ANN models have a similar level of accuracy . It was concluded that for the data series studied the ANN models show a more appropriate Ibovespa forecasting than the traditional models of time series, represented by the GARCH model