Superiority Of Artificial Neural Networks Research Articles

Superiority of artificial neural networks over conventional hydrological models in simulating urban catchment runoff

ABSTRACT The synergistic impacts of climate change and urbanisation have amplified the recurrence and austerity of intense rainfall events, exacerbating persistent flooding risk in urban environments. The intricate topography and inherent non-linearity of urban hydrological processes limit the predictive accuracy of conventional models, leading to significant discrepancies in flow estimation. Recent advancements in artificial neural network (ANNs) have demonstrated remarkable progress in mitigating most limitations, specifically in simulating complex, non-linear relationships, without an intricate comprehension of the underlying physical processes. This paper proposes a deep learning ANN-based flow estimation model for enhanced precision simulation of streamflow in urban catchments, with the research's distinctive contribution involving rigorous comparative evaluation of the developed model against the established Australian hydrological model, RORB. Gardiners Creek catchment, an urban catchment situated in East Melbourne was designated as the study area, with the model being calibrated upon historical storm incidences. The findings reveal that the ANN model substantially outperforms RORB, as evidenced by superior correlation, prediction efficiency, and lower generalisation error. This underscores the ANN's adeptness in accurately replicating non-linear-catchment responses to storm events, marking a substantial advancement over conventional modelling practices and indicating its transformative potential for enhancing flood prediction precision and revolutionising current estimation practices.

Modeling Carbon Release of Brazilian Highest Economic Pole and Major Urban Emitter: Comparing Classical Methods and Artificial Neural Networks

Despite the concern about climate change and the associated negative impacts, fossil fuels continue to prevail in the global energy consumption. This paper aimed to propose the first model that relates CO2 emissions of Sao Paulo, the main urban center emitter in Brazil, with gross national product and energy consumption. Thus, we investigated the accuracy of three different methods: multivariate linear regression, elastic-net regression, and multilayer perceptron artificial neural networks. Comparing the results, we clearly demonstrated the superiority of artificial neural networks when compared with the other models. They presented better results of mean absolute percentage error (MAPE = 0.76%) and the highest possible coefficient of determination (R2 = 1.00). This investigation provides an innovative integrated climate-economic approach for the accurate prediction of carbon emissions. Therefore, it can be considered as a potential valuable decision-support tool for policymakers to design and implement effective environmental policies.

PREVISÃO DO PERFIL DAS INSTITUIÇÕES ENVOLVIDAS EM ESTRATÉGIAS DE FUSÕES E AQUISIÇÕES (F&A) DO SETOR BANCÁRIO BRASILEIRO

O objetivo deste trabalho foi identificar a importância de variáveis econômico-financeiras para a ocorrência de fusões e aquisições (F&A) no setor bancário brasileiro após 20 anos consolidados do plano-real, período compreendido entre os anos de 1995 e 2015. Para atingir o objetivo proposto, foram utilizadas as técnicas de análise discriminante, regressão logística, redes neurais e um modelo híbrido. De maneira geral observou-se que os indicadores de qualidade dos ativos, rentabilidade, liquidez, eficiência e tamanho da firma foram importantes na discriminação dos grupos de bancos estudados (adquirentes e adquiridos) e foi possível constatar que bancos com maiores indicadores apresentam maior probabilidade se tornarem adquirentes. Com relação aos métodos empregados pode-se afirmar que os modelos apresentaram aderência aos dados estudados, todavia, ressalta-se a superioridade das redes neurais artificiais em sua forma tradicional e híbrida. Por fim, salienta-se a importância de trabalhos como este em mercados emergentes, sendo que modelos de previsão podem trazer mais segurança e amenizar os riscos assumidos pelos investidores. Além disso, fornecem informações úteis para a tomada de decisão empresarial, uma vez que elencam variáveis importantes para a classificação de empresas alvo e não alvo de F&A.

Intelligent forecasting with machine learning trading systems in chaotic intraday Bitcoin market

Abstract Due to the remarkable boost in cryptocurrency trading on digital blockchain platforms, the utilization of advanced machine learning systems for robust prediction of highly nonlinear and noisy data, gains further popularity by individual and institutional market agents. The purpose of our study is to comparatively evaluate a plethora of Artificial Intelligence systems in forecasting high frequency Bitcoin price series. We employ three different sets of models, i.e., statistical machine learning approaches including support vector regressions (SVR) and Gaussian Poisson regressions (GRP), algorithmic models such as regression trees (RT) and the k-nearest neighbours (kNN) and finally artificial neural network topologies such as feedforward (FFNN), Bayesian regularization (BRNN) and radial basis function networks (RBFNN). To the best of our knowledge, this is the first time an extensive empirical investigation of the comparative predictability of various machine learning models is implemented in high-frequency trading of Bitcoin. The entropy analysis of training and testing samples reveals long memory traits, high levels of stochasticity, and topological complexity. The presence of inherent nonlinear dynamics of Bitcoin time series fully rationalizes the use of advanced machines learning techniques. The optimal parameter values for SVR, GRP and kNN are found via Bayesian optimization. Based on diverse performance metrics, our results show that the BRNN renders an outstanding accuracy in forecasting, while its convergence is unhindered and remarkably fast. The overall superiority of artificial neural networks is due to parallel processing features that efficiently emulate human decision-making in the presence of underlying nonlinear input-output relationships in noisy signal environments.

Effective Z evaluation using monoenergetic gamma rays and neural networks

Two analysis methods for $$\textit{Z}_\mathrm{eff}$$ evaluation were explored using both experimental and simulated gamma-ray attenuation data. Using particle-capture reactions on composite targets to generate multi-monoenergetic gamma rays between 1 and 12 MeV, we demonstrate the advantage of using neural networks for effective Z evaluation of shielded materials in single-pixel measurements. Furthermore, we extend the analysis to 2D processing of transmission radiography and by using Geant4-simulated data we prove the superiority of artificial neural networks in terms of image quality and material discrimination against classical methods.

Data-driven heating and cooling load predictions for non-residential buildings based on support vector machine regression and NARX Recurrent Neural Network: A comparative study on district scale

Predicting building energy consumption is essential for planning and managing energy systems. In recent times, numerous studies focus on load forecasting models dealing with a wide range of different methods. In addition to Artificial Neural Networks (ANN), especially Support Vector Machines (SVM) have been studied. Various research work showed the success and superiority of ANN and SVM for load predictions, where frequently, SVM outperformed ANN models. In this study, data-driven thermal load forecasting performance of ε-SVM Regression (ε-SVM-R) based on a Radial Basis Function (RBF) and a polynomial kernel is compared to the outcome of two Nonlinear Autoregressive Exogenous Recurrent Neural Networks (NARX RNN) of different depths. For demonstration, historical data from a non-residential district in Germany is used for training and testing to predict monthly loads. The evaluation of the resulting predictions show that NARX RNNs yields higher accuracy than (ε-SVM-R) models, in combination with comparable computational effort.

Scene Text Recognition using Artificial Neural Network: A Survey

scene text recognition has become an important emerging area of research in the field of image processing. In image processing, character recognition boosts the complexity in the area of Artificial Intelligence. Character recognition is not easy for computer programs in comparison to humans. In the broad spectrum of things, it may consider that recognizing patterns is the only thing which humans can do well and computers cannot. There are many reasons including various sources of variability, hypothesis and absence of hard-and-fast rules that define the appearance of a visual character. Hence; there is an unavoidable requirement for heuristic deduction of rules from different samples. This review highlights the superiority of artificial neural networks, a popular area of Artificial Intelligence, over various other available methods like fuzzy logic and genetic algorithm. In this paper, two methods are listed for character recognition - offline and online. The ―Offline‖ methods include Feature Extraction, Clustering, and Pattern Matching. Artificial neural networks use the static image properties. The online methods are divided into two methods, k-NN classifier and direction based algorithm. Thus, the scale of techniques available for scene text recognition deserves an admiration. This review gives a detail survey of use of artificial neural network in scene text recognition. KeywordsRecognition, Scene text recognition,Text extraction,

Superiority of artificial neural networks for a genetic classification procedure.

The correct classification of individuals is extremely important for the preservation of genetic variability and for maximization of yield in breeding programs using phenotypic traits and genetic markers. The Fisher and Anderson discriminant functions are commonly used multivariate statistical techniques for these situations, which allow for the allocation of an initially unknown individual to predefined groups. However, for higher levels of similarity, such as those found in backcrossed populations, these methods have proven to be inefficient. Recently, much research has been devoted to developing a new paradigm of computing known as artificial neural networks (ANNs), which can be used to solve many statistical problems, including classification problems. The aim of this study was to evaluate the feasibility of ANNs as an evaluation technique of genetic diversity by comparing their performance with that of traditional methods. The discriminant functions were equally ineffective in discriminating the populations, with error rates of 23-82%, thereby preventing the correct discrimination of individuals between populations. The ANN was effective in classifying populations with low and high differentiation, such as those derived from a genetic design established from backcrosses, even in cases of low differentiation of the data sets. The ANN appears to be a promising technique to solve classification problems, since the number of individuals classified incorrectly by the ANN was always lower than that of the discriminant functions. We envisage the potential relevant application of this improved procedure in the genomic classification of markers to distinguish between breeds and accessions.

Correct vs. accurate prediction: A comparison between prediction power of artificial neural networks and logistic regression in psychological researches

This is a comparative research, which was carried out to study the abilities of two prediction models in psychological researches; Logistic Regression (LR) versus Artificial Neural Networks (ANNs). Four hundred fifty six students were chosen randomly from one of the educational areas in Tehran (Iran). Eighteen psychological traits and five levels of adjustment were considered as predictor and predicted variables, respectively. According to the first assessment, the ANNs were more successful than LR. By reduction of the adjustment levels from five to three, this superiority of ANNs was changed in the favour of LR. So there are two definitions for the power of prediction: one refers to the correctness and the other to the accuracy of prediction.

Quantitative Structure-Retention Relationship Study of Some Phenol Derivatives in Gas Chromatography

A quantitative structure-retention relationship (QSRR) model has been developed for the gas chromatographic retention times of 37 phenolic derivatives in a DB-5 non-polar column (95% dimethyl and 5% diphenyl-polysiloxane). As a first step, multiple linear regression (MLR) was employed to gain informative descriptors that can predict the retention times of these compounds. Descriptors appearing in the MLR model are categorized as topological and geometric parameters that comply with the applied column. Furthermore, each molecular descriptor in this model was examined to unfold the relationship between molecular structures and their retention times. Then, a 4-4-1 neural network was developed using the descriptors selected by the MLR model. The comparison of the standard errors and correlation coefficients reveals the superiority of artificial neural networks (ANN) over the MLR model. This refers to the fact that the retention behaviors of molecules display non-linear characteristics. The consistency and reliability of ANN model was investigated using the L4O cross-validation technique. The obtained results are closely in compliance with the experiment. Moreover, the mean effect of descriptors shows that Kier symmetry index is the most important factor affecting the retention behavior of molecules.

Usefulness of artificial neural networks to predict follow-up dietary protein intake in hemodialysis patients

Artificial neural networks (ANN) represent a promising alternative to classical statistical and mathematic methods to solve multidimensional nonlinear problems. The aim of the study was to verify, by comparing the performance of ANN with that of experienced nephrologists, whether ANN are useful tools in hemodialysis to predict the follow-up (=1 month after the observation used for the prediction) dietary protein intake (PCR), and whether their performance is influenced by the size of the population and by the data pool used to built the model. A combined retrospective and prospective observational study was performed in two Swiss dialysis units (84 chronic hemodialysis patients, 500 monthly clinical observations and biochemical test results). Using mathematical models based on linear regressions to evaluate the variables, ANN were built and then prospectively and interinstitutionally compared with the ability of six experienced nephrologists to predict the follow-up PCR. ANN compared with nephrologists gave a more accurate correlation between estimated and calculated follow-up PCR (P < 0.001). The same superiority of ANN was also seen in the ability to detect a follow-up PCR <1.00 g/kg/day expressed as a percentage of correct predictions, sensitivity, specificity, and predictivity. The interinstitutional performance of the ANN is positively influenced by the size and the variability of the population used to build the mathematical model. The use of ANN significantly improves the ability of the experienced nephrologist to estimate and to detect an unsatisfactory (<1.00 g/kg/day) follow-up PCR. The size of the population selected to build the ANN is critical for his performance.

Artificial neural networks improve the prediction of Kt/V, follow-up dietary protein intake and hypotension risk in haemodialysis patients.

Artificial neural networks (ANN) represent a promising alternative to classical statistical and mathematical methods to solve multidimensional non-linear problems. The aim of the study was to compare the performance of ANN in predicting the dialysis quality (Kt/V), the follow-up dietary protein intake and the risk of intradialytic hypotension in haemodialysis patients with that predicted by experienced nephrologists. A combined retrospective and prospective observational study was performed in two Swiss dialysis units (80 chronic haemodialysis patients, 480 monthly clinical observations and biochemical test results). Using mathematical models based on linear and logistic regressions as background, ANN were built and then prospectively compared with the ability of six experienced nephrologists to predict the Kt/V and the follow-up protein catabolic rate (PCR) and to detect a Kt/V < 1.30, a follow-up PCR < 1.00 g/kg/day and the occurrence of hypotension. ANN compared with nephrologists gave a more accurate correlation between estimated and calculated Kt/V and follow-up PCR (P<0.001). The same superiority of ANN was also seen in the ability to detect a Kt/V < 1.30, a follow-up PCR < 1.00 g/kg/day and the occurrence of hypotension expressed as a percentage of correct answers, sensitivity, specificity and predictivity. The use of ANN significantly improves the ability of experienced nephrologists to estimate the Kt/V and the follow-up PCR and to detect a Kt/V < 1.30, a follow-up PCR < 1.00 g/kg/day and the occurrence of intradialytic hypotension.

Artificial Neural Networks–Modern Systems for Safety Control

A short review of the applications of artificial neural networks in different fields of industry with a description of their main properties is made. Such systems have specific properties typical for the human brain, which can decide on the superiority of artificial neural networks over standard control systems. Basic types of such networks as well as their principles of operation and successful applications are described. The application of artificial neural networks in safety engineering is discussed with stress on their special properties, which are necessary in safety critical systems.