Traditional Back Propagation Neural Network Research Articles

Study of carbon dioxide emissions prediction in Hebei province, China using a BPNN based on GA

With the deterioration of the global greenhouse effect, the study of carbon dioxide emissions has received more and more international attention and accurate prediction of carbon dioxide emissions is also important for the formulation of reasonable energy-saving emission reduction measures. In this paper, the genetic algorithm is used to optimize the initial connection weights and thresholds of the traditional back propagation neural network (BPNN) which can give full play to the advantages of the genetic algorithm's global search capacity and BPNN's local search. The data of Hebei province in China during the period 1978–2012 are selected to carry out the carbon dioxide emissions prediction with the established model. In the view of the choice of input variables, the coal consumption, gross domestic product, total population, and urbanization level are examined by Pearson coefficient test. Auto correlation and partial correlation are applied to analyze the inner relationships between the historic carbon dioxide emissions, thus to select the input variables of BPNN. Besides, in order to verify the validity of the built model, the residual auto correlation and partial correlation are done upon the training set. The prediction results suggest the proposed model outperforms the compared models.

Financial security evaluation of the electric power industry in China based on a back propagation neural network optimized by genetic algorithm

Recently security issues like investment and financing in China's power industry have become increasingly prominent, bringing serious challenges to the financial security of the domestic power industry. Thus, it deserves to develop financial safety evaluation towards the Chinese power industry and is of practical significance. In this paper, the GA (genetic algorithm) is used to optimize the connection weights and thresholds of the traditional BPNN (back propagation neural network) so the new model of BPNN based on GA is established, hereinafter referred to as GA-BPNN (back propagation neural network based on genetic algorithm). Then, an empirical example of the electric power industry in China during the period 2003–2010 was selected to verify the proposed algorithm. By comparison with three other algorithms, the results indicate the model can be applied to evaluate the financial security of China's power industry effectively. Then index values of the financial security of China's power industry in 2011 were obtained according to the tested prediction model and the comprehensive safety scores and grades are calculated by the weighted algorithm. Finally, we analyzed the reasons and throw out suggestions based on the results. The work of this paper will provide a reference for the financial security evaluation of the energy industry in the future.

Application of genetic algorithm-back propagation for prediction of mercury speciation in combustion flue gas

Coal combustion is one of the main sources of mercury emission. Studies using artificial neural networks (ANNs) to predict mercury emission have shown the feasibility of ANN method. Such analyses aimed to provide guidance for mercury emission control in coal combustion. A mercury emission prediction model was developed by modifying the traditional back propagation (BP) neural networks, and a genetic algorithm (GA) based on global search was used, so called the GA-BP neural networks. In total, six main factors were evaluated and selected as the characteristics parameters. Totally, 20 coal-fired boilers were used as training samples, and three different types of mercury including elemental mercury, oxidized mercury, and particulate mercury were used as outputs. The accuracy of prediction results was analyzed, and source of error was discussed. Results show that correlation efficiency for the training samples was as high as 0.895. Three additional samples were studied to test the predictive model. Results of training and predicting were highly correlated with actual measurement results. It is shown that GA-BP is a promising model for mercury speciation prediction.

Epileptic seizure detection based on improved wavelet neural networks in long-term intracranial EEG

Automatic seizure detection is of great importance for speeding up the inspection process and relieving the workload of medical staff in the analysis of EEG recordings. In this study, a method based on an improved wavelet neural network (WNN) is proposed for automatic seizure detection in long-term intracranial EEG. WNN combines the traditional back propagation neural network (BPNN) with wavelet transform. Compared with classic WNN architectures, a modified point symmetry-based fuzzy c-means (MSFCM) algorithm is applied to the initialization of wavelet transform's translations, which has been successful in multiclass cancer classification. In addition, Fast-decaying Morlet wavelet is chosen as the activation function to make the WNN learn faster. Relative amplitude and relative fluctuation index are extracted as a feature vector to describe the variation of EEG signals, and the feature vector is then fed into WNN for classification. At last, post-processing including smoothing, channel fusion and collar technique is adopted to achieve more accurate and stable results. This system performs efficiently with the average sensitivity of 96.72%, specificity of 98.91% and false-detection rate of 0.27h−1. The proposed approach achieves high sensitivity and low false detection rate, which demonstrates its potential for clinical usage.

Multi-sensor data fusion by a hybrid methodology – A comparative study

Multi-sensor data fusion is considered as an inherent problem in wireless sensor network applications. It is widely assumed as a sturdy non linear system in view of the complexities involved in its operation. An accurate and precise methodical solution is therefore a complicated task to accomplish. It is crucial for the sensory systems that they should not be influenced in terms of accuracy and precision by any means. To address these issues a hybrid model employing rough set (RS) with back-propagation neural network (BPNN) is used to ameliorate the data fusion capability of the system with an illustrative example. Experimental results have demonstrated an escalating improvement in the predictive accuracy of the hybrid model as compared to the traditional BPNN model.

Robust modelling and prediction of thermally induced positional error based on grey rough set theory and neural networks

Thermal errors adversely affect the precision of a machine tool operation. Therefore, reduction of thermal errors is essential for improving of machining accuracy. In this paper, in order to realize real-time compensation, taking a multi-axis machine tool as an example and a method for robust modeling and predicting thermally induced positional error is proposed. First of all, the number of temperature measuring points required in thermal error’s model was reduced based on the rough set theory, which greatly reduced variable searching and modeling time. Then through grey relation theory, systematic analysis of the similarity degree between thermal error and temperature data was carried out to select sensitive temperature measuring points, and the temperature variables in the thermal error’s model were reduced from 24 to 7 after optimization, which eliminated the coupling problems. For reducing the influence of unpredictable noises, radial basis function (RBF) and back propagation (BP) neural network modeling methods were adopted to predict thermally induced positional errors, and in comparison, the prediction accuracy of RBF neural network was found superior to that of traditional BP neural network. Finally, some measured data were selected to verify the validity of the proposed method, and the results showed that prediction accuracy of the proposed thermally induced positional error model was reliable.

Using a back propagation neural network based on improved particle swarm optimization to study the influential factors of carbon dioxide emissions in Hebei Province, China

The emission of carbon dioxide is the primary cause of the greenhouse effect, therefore a precise study of the influential factors of carbon dioxide emissions is of great significance to control the growth from the source. In this paper, non-inertial weight coefficients and selective mutation strategies are used in a particle swarm optimization algorithm, and the improved particle swarm was used to optimize the initial connection weights and thresholds of a traditional back propagation (BP) neural network. Consequently, a new BP model based on an improved particle swarm (IPSO) is established: improved particle swarm optimization-back propagation algorithm (IPSO-BP). In order to verify the overall performance and effectiveness of the proposed method, an empirical analysis of carbon dioxide emissions and influential factors was carried out in Hebei Province, China during the period 1978–2012. The results were compared with those of two other methods to prove that the proposed IPSO-BP algorithm could take full advantage of IPSO's global search capability and BP's local search capability, as well as overcome the problems of BP of random initial values and premature solutions. In addition, the precision of the fit and prediction of carbon dioxide emissions are improved notably.

Smart Information Reconstruction via Time-Space-Spectrum Continuum for Cloud Removal in Satellite Images

Cloud contamination is a big obstacle when processing satellite images retrieved from visible and infrared spectral ranges for application. Although computational techniques including interpolation and substitution have been applied to recover missing information caused by cloud contamination, these algorithms are subject to many limitations. In this paper, a novel smart information reconstruction (SMIR) method is proposed, in order to reconstruct cloud contaminated pixel values from the time-space-spectrum continuum with the aid of a machine learning tool, namely extreme learning machine (ELM). For the purpose of demonstration, the performance of SMIR is evaluated by reconstructing the missing remote sensing reflectance values derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) on board the Terra satellite over Lake Nicaragua, where is a very cloudy area year round. For comparison, the traditional backpropagation neural network algorithms will also be implemented to reconstruct the missing values. Experimental results show that the ELM outperforms the BP algorithms by an enhanced machine learning capacity with simulated memory effect embedded in MODIS due to linking the complex time-space-spectrum continuum between cloud-free and cloudy pixels. The ELM-based SMIR practice presents a correlation coefficient of 0.88 with root mean squared error of 7.4E - 04sr <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">-1</sup> between simulated and observed reflectance values. Finding suggests that the SMIR method is effective to reconstruct all the missing information providing visually logical and quantitatively assured images for further image processing and interpretation in environmental applications.

Self-adaptive extreme learning machine

In order to overcome the disadvantage of the traditional algorithm for SLFN (single-hidden layer feedforward neural network), an improved algorithm for SLFN, called extreme learning machine (ELM), is proposed by Huang et al. However, ELM is sensitive to the neuron number in hidden layer and its selection is a difficult-to-solve problem. In this paper, a self-adaptive mechanism is introduced into the ELM. Herein, a new variant of ELM, called self-adaptive extreme learning machine (SaELM), is proposed. SaELM is a self-adaptive learning algorithm that can always select the best neuron number in hidden layer to form the neural networks. There is no need to adjust any parameters in the training process. In order to prove the performance of the SaELM, it is used to solve the Italian wine and iris classification problems. Through the comparisons between SaELM and the traditional back propagation, basic ELM and general regression neural network, the results have proven that SaELM has a faster learning speed and better generalization performance when solving the classification problem.

An optimized classification algorithm by BP neural network based on PLS and HCA

Due to some correlative or repetitive factors between features or samples with high dimension and large amount of sample data, when traditional back-propagation (BP) neural network is used to solve this classification problem, it will present a series of problems such as network structural redundancy, low learning efficiency, occupation of storage space, consumption of computing time, and so on. All of these problems will restrict the operating efficiency and classification precision of neural network. To avoid them, partial least squares (PLS) algorithm is used to reduce the feature dimension of original data into low-dimensional data as the input of BP neural network, so that it can simplify the structure and accelerate convergence, thus improving the training speed and operating efficiency. In order to improve the classification precision of BP neural network by using hierarchical cluster analysis (HCA), similar samples are put into a sub-class, and some different sub-classes can be obtained. For each sub-class, a different training session can be conducted to find a corresponding precision BP neural network model, and the simulation samples of different sub-classes can be recognized by the corresponding network model. In this paper, the theories of PLS and HCA are combined together with the property of BP neural network, and an optimized classification algorithm by BP neural network based on PLS and HCA (PLS-HCA-BP algorithm) is proposed. The new algorithm is aimed at improving the operating efficiency and classification precision so as to provide a more reliable and more convenient tool for complex pattern classification systems. Three experiments and comparisons with four other algorithms are carried out to verify the superiority of the proposed algorithm, and the results indicate a good picture of the PLS-HCA-BP algorithm, which is worthy of further promotion.

Forecasting stock market indexes using principle component analysis and stochastic time effective neural networks

Financial market dynamics forecasting has long been a focus of economic research. A stochastic time effective function neural network (STNN) with principal component analysis (PCA) developed for financial time series prediction is presented in the present work. In the training modeling, we first use the approach of PCA to extract the principal components from the input data, then integrate the STNN model to perform the financial price series prediction. By taking the proposed model compared with the traditional backpropagation neural network (BPNN), PCA-BPNN and STNN, the empirical analysis shows that the forecasting results of the proposed neural network display a better performance in financial time series forecasting. Further, the empirical research is performed in testing the predictive effects of SSE, HS300, S&P500 and DJIA in the established model, and the corresponding statistical comparisons of the above market indices are also exhibited.

Deformation Monitoring Model of Jinping-I Hydropower Station Based on NACA-BP Neural Network

Back Propagation (BP) neural network can learn and store a large number of input-output model nonlinear relationships with simple structure. Niche ant colony algorithm (NACA) combines the ant colony algorithm (ACA) with the niche technology in order to add its local search ability to ACA with preserving the intelligent search ability and robustness of ACA. To optimize predicting model establishment of the dam monitoring data, NACA and BP neural network modeling method are combined to establish a prediction model of horizontal displacement monitoring data. The traditional BP neural network prediction model is established to make a comparison with the NACA. The results show that NACA-BP neural network method can speed up the convergence rate of BP neural network and enhance local search ability and prediction accuracy.

Sine neural network (SNN) with double-stage weights and structure determination (DS-WASD)

To solve complex problems such as multi-input function approximation by using neural networks and to overcome the inherent defects of traditional back-propagation neural networks, a single hidden-layer feed-forward sine-activated neural network, sine neural network (SNN), is proposed and investigated in this paper. Then, a double-stage weights and structure determination (DS-WASD) method, which is based on the weights direct determination method and the approximation theory of using linearly independent functions, is developed to train the proposed SNN. Such a DS-WASD method can efficiently and automatically obtain the relatively optimal SNN structure. Numerical results illustrate the validity and efficacy of the SNN model and the DS-WASD method. That is, the proposed SNN model equipped with the DS-WASD method has great performance of approximation on multi-input function data.

The Applications of BP Neural Network Based on MIV in Hydraulic Fracturing

As a effective stimulation, hydraulic fracturing was commonly used in the mining process of complicated low permeability reservoirs, especially in horizontal shale gas wells more widely. In this paper, we establish the BP (Back Propagation) neural network based on MIV (Mean Impact Value) which is different from traditional BP neural network. We choose the independent variables of training set, plus or minus a certain percentages. Measured by the absolute size of MIV value and analyze the changes of MIV value, we identify the main factors in hydraulic fracturing. Combined qualitative analysis with quantitative calculation, the hydraulic fracturing evaluation model was established. Compared with the traditional BP neural network, the evaluation model shows a high precision, and good stability. Above all, we succeed in achieving a rapid, objective and impartial productivity prediction after hydraulic fracturing in complicated low permeability reservoirs. So in the future, this new hydraulic fracturing evaluation model is about to having abroad applications in the field of hydraulic fracturing.

Correlation Coefficient Matrix Based PD Feature Dimensional Reduction

The abstraction of diagnostic feature from field condition monitoring data is a significant research challenge. A new dimension reduction method based on correlation coefficient matrix is proposed aimed at the high-dimension characteristic parameters in the process of pattern recognition for partial discharge in power transformer. The CCM is constructed by parameters extracted from partial discharge signature in power transformer. The parameters that have similar classification characters are reduced directed by the correlation analysis result. The reduced PD features are inputted to the pattern classifiers of probabilistic neural networks (PNN). The results show that the parameter dimension is reduced and the classifier construction is simplified, and the recognition effect is better than that of the traditional back propagation neural network (BPNN) in the condition of small samples.

Natural Gas Load Forecasting Based on Improved Back Propagation Neural Network

To suit for the condition that the relative error is more popular than the absolute error, and overcome the shortcoming of the traditional Back propagation neural network, this paper proposed an improved Back propagation algorithm with additional momentum item based on the sum of relative error square. The improved algorithm was applied to the example of the natural gas load forecasting, simulations showed that the improved algorithm has faster training speed than the traditional algorithm, and has higher accuracy as while.

An Optimized Neural Network Classifier for Automatic Modulation Recognition

Automatic modulation recognition which is one of the key technologies in no-cooperative communications has extensive application prospects in civilian and military fields. The design of classifier played a decisive role in recognition results. The classifier based on back propagation (BP) neural network is better in the existing methods. However, the traditional back propagation neural network (BPNN) have some well-known disadvantages. This study investigates the design of a classifier for recognition of six common digital modulations. This classifier based on BP neural network trained by improved particle swarm optimization (PSO) which is applied as a local search algorithm to find the optimal weights and thresholds of BPNN. The simulation experiment results demonstrate that the proposed classifier has a higher recognition accuracy than other classifiers. DOI : http://dx.doi.org/10.11591/telkomnika.v12i2.3930

A New Technique to Improve Estimation of Position for Serial Robots

This paper presents a new method to improve the estimation of the positions for serial robots using power-activated feed-forward neural network. In the paper, a six-input three-output neural network is created with robot joint angle sine values as inputs and positions in the world frame as outputs. The neuron is activated with an orthogonal polynomial sequence,and the neural weights can be calculated directly without involving iterative and convergent problem. It is found that, the RMS error is less than 0.25 mm for the whole work space. And the absolute and relative errors of this method are smaller than those of built in kinematics model and the traditional back propagation (BP) neural network method. Experimental results show that the proposed method can effectively predict the positioning of the given joint angles.

Thermal error modeling by integrating GA and BP algorithms for the high-speed spindle

Thermal error is one of primary reasons affecting the cutting accuracy of a machine tool. Especially, high-speed spindle’s thermal error restricts the improvement of high-precision machine tool accuracy. In this paper, the traditional back propagation (BP) neural network modeling principle is firstly analyzed. Then, considering the BP network, which is simple and adaptive but converges slowly and is easy to reach local minima, a genetic algorithm (GA) is introduced to optimize BP network’s initial weights and thresholds. Lastly, the network is trained using combined GA and BP technologies with practical thermal error sample data. In experiments of thermal error prediction of the high-speed spindle in a machine tool, the average compensation rate is increased from 89.03 % of BP model to 93.155 % of GABP model. Therefore, GABP model shows its effectiveness in quickly solving the global minimum searching problem.

An Evaluation Model for Tailings Storage Facilities Using Improved Neural Networks and Fuzzy Mathematics

With the development of mine industry, tailings storage facility (TSF), as the important facility of mining, has attracted increasing attention for its safety problems. However, the problems of low accuracy and slow operation rate often occur in current TSF safety evaluation models. This paper establishes a reasonable TSF safety evaluation index system and puts forward a new TSF safety evaluation model by combining the theories for the analytic hierarchy process (AHP) and improved back-propagation (BP) neural network algorithm. The varying proportions of cross validation were calculated, demonstrating that this method has better evaluation performance with higher learning efficiency and faster convergence speed and avoids the oscillation in the training process in traditional BP neural network method and other primary neural network methods. The entire analysis shows the combination of the two methods increases the accuracy and reliability of the safety evaluation, and it can be well applied in the TSF safety evaluation.