Combined Neural Network Model Research Articles

Beamforming and lightweight GRU neural network combination model for multi-channel speech enhancement

Beamforming and lightweight GRU neural network combination model for multi-channel speech enhancement

Prediction of Longitudinal Resonance Elastic Modulus of Wood-plastic Composites Based on GM (1, N) Neural Network

Abstract Based on the grey neural network combined prediction model, the prediction model of the mechanical properties of WPC composites was established by taking the orthogonal design test results of the main processing factors of WPC composites as data samples. Taking the main processing factors of WPC as input, grey GM (1, N) prediction of these variables was carried out, and then BP neural network prediction was carried out to predict the mechanical process index of WPC. The results show that the established prediction model can predict the mechanical properties and technological indexes of WPC composites well.

Development of an Emulator for Radiation Physics Simulator Using Stochastic Variational Gaussian Process Model

This paper describes an emulator that uses a Stochastic Variational Gaussian process (SVGP) regression model to parameterize radiation in a numerical weather prediction (NWP) model that meteorologically models the Earth's weather system. The computation of radiative processes is very large, accounting for most of the total NWP model computation. Statistical emulators are surrogate models that represent simulators and can overcome the computational limitations of very complex simulators such as radiative processes. Recently, artificial neural network-based radiative transfer emulators have been developed, and in this study, a statistical model, GP, is used to develop a radiative transfer emulator. The GP model has the advantage of calculating the uncertainty of the prediction along with the prediction, so the uncertainty of the prediction can be utilized appropriately. However, the computational complexity of the conventional GP model is very high, making it difficult to apply to large data. To solve this problem, an approximate approach, the SVGP model, was utilized. To further reduce the dimensionality of the input variables, we used a combined neural network and SVGP model. As a result, the SVGP-based radiative physics emulator improved its accuracy by about 20% compared to the artificial neural network emulator. However, the computation speed was about 3 to 9 times slower than the neural network emulator, but it was faster than the computation speed of the NWP model. This suggests that statistical emulators can be used to replace NWP model simulators.

Hospital Management Practice of Combined Prediction Method Based on Neural Network

In this article, the outpatient volume, hospitalization income and drug demand in hospital management are taken as the research objects, and a neural network combined prediction model is established to predict the outpatient volume with the fitting prediction results of cubic polynomial regression model and grey model as the input of the network and the actual statistical outpatient volume as the output. Lasso variable selection method is used to determine the main indexes affecting the income of inpatients in hospital, and a prediction model combining grey prediction and artificial neural network is established to predict the income of inpatients in hospital. By studying the key characteristics of hospital drug demand, BP neural network, RBF neural network and GRNN generalized regression neural network are selected to predict the drug demand. By solving the quadratic programming problem and according to the weight rules, a combination forecasting model based on neural network is established to predict the drug demand, and the accuracy and stability of the model are evaluated.

Interpretable zero-inflated neural network models for predicting admission counts

Abstract In this paper, we construct interpretable zero-inflated neural network models for modeling hospital admission counts related to respiratory diseases among a health-insured population and their dependants in the United States. In particular, we exemplify our approach by considering the zero-inflated Poisson neural network (ZIPNN), and we follow the combined actuarial neural network (CANN) approach for developing zero-inflated combined actuarial neural network (ZIPCANN) models for modeling admission rates, which can accommodate the excess zero nature of admission counts data. Furthermore, we adopt the LocalGLMnet approach (Richman & Wüthrich (2023). Scandinavian Actuarial Journal, 2023(1), 71–95.) for interpreting the ZIPNN model results. This facilitates the analysis of the impact of a number of socio-demographic factors on the admission rates related to respiratory disease while benefiting from an improved predictive performance. The real-life utility of the methodologies developed as part of this work lies in the fact that they facilitate accurate rate setting, in addition to offering the potential to inform health interventions.

An improved sparrow search algorithm and CNN-BiLSTM neural network for predicting sea level height

Accurate prediction of sea level height is critically important for the government in assessing sea level risk in coastal areas. However, due to the nonlinear, time-varying and highly uncertain characteristics of sea level change data, sea level prediction is challenging. To improve the accuracy of sea level prediction, this paper uses a new swarm intelligence algorithm named the sparrow search algorithm (SSA), which can imitate the foraging behavior and antipredation behavior of sparrows, to determine optimal solutions. To avoid the algorithm falling into a local optimal situation, this paper integrates the sine–cosine algorithm and the Cauchy variation strategy into the SSA to obtain an algorithm named the SCSSA. The SCSSA is used to optimize the parameter values of the CNN-BiLSTM (convolutional neural network combined with bidirectional long short-term memory neural network) model; finally, a combined neural network model (named SCSSA-CNN-BiLSTM) is proposed. In this paper, the time series data of seven tidal stations located in coastal China are used for experimental analysis. First, the SCSSA-CNN-BiLSTM model is compared with the CNN-BiLSTM model to predict the time series data of SHANWEI Station. With respect to the training and test sets of data, the SCSSA-CNN-BiLSTM model outperforms the other models on all the evaluation metrics. In addition, the remaining six tide station datasets and five neural network models, including the SCSSA-CNN-BiLSTM model, are used to further study the performance of the proposed prediction model. Four evaluation indices including the root mean squared error (RMSE), mean absolute error (MAE), mean absolute percentage error (MAPE) and coefficient of determination (R2) are adopted. For six stations, the RMSE, MAE, MAPE and R2 of SCSSA-CNN-BiLSTM model are ranged from 20.9217 ~ 27.8427 mm, 9.4770 ~ 17.8603 mm, 0.1322% ~ 0.2482% and 0.9119 ~ 0.9759, respectively. The experimental analysis results show that the SCSSA-CNN-BiLSTM model makes effective predictions at all stations, and the prediction performance is better than that of the other models. Even though the combination of SCSSA algorithm may increase the complexity of the model, indeed the proposed model is a new prediction method with good accuracy and robustness for predicting sea level change.

Research on fault diagnosis method of linear vibration screen based on fused RP-improved CNN

A fault diagnosis method based on fused RP (recurrence plot) and improved CNN (convolutional neural networks) is proposed for the traditional fault diagnosis method of linear vibrating screen manually designed and optimized features with feature quality uncertainty. A deep convolutional neural network combination model (MDCNN) with high-level feature fusion is designed. The collected multi-source vibration signals were converted into black and white recurrence plot, and fusion into three-channel RGB recurrence plot. Fault diagnosis was classified by the combination model of deep convolution neural network with high level feature fusion. The results show that the recurrence plot of homologous multi-sensor signals contains more information about the characteristics of the vibrating screen than the original colorful signal map or the Gram Point Field, which contributes to the feature learning and classification of MDCNN, with an average recognition accuracy of 97.59% under strong vibrating sift-5db noise. Compared with manifold classical neural network models, the MDCNN model is proved to be effective, stable and small space occupying.

Demand and supply gap analysis of Chinese new energy vehicle charging infrastructure: Based on CNN-LSTM prediction model

The sales of new energy vehicles (NEVs) and the construction of charging infrastructure promote and constrain each other. It is crucial for the development of the new energy vehicle industry to understand the gap clearly and accurately between the supply and demand of NEV charging infrastructure. In this paper, a neural network combined model based on convolutional neural network (CNN) and long and short-term memory (LSTM) is introduced for accurate prediction of NEVS sales and charging infrastructure ownership. Compared with other traditional and combined models, the CNN-LSTM combined model performs best in multiple evaluation metrics while using less computing power. The RMSE, MAE, MAPE, and R2 of the CNN-LSTM combined model were 52.80, 42.67, 17 %, and 0.78, respectively. Accordingly, it is sufficient to demonstrate the excellent prediction performance of the CNN-LSTM combined model constructed in this paper. The forecast results show that in 2025, the ratio of NEVs to public charging piles will rise to 10.2:1 and the ratio to private charging piles will fall to 2.5:1. The overall ratio shows a downward trend and is expected to reach 2:1. There is a gap in the demand for NEV charging infrastructure. Finally, this paper makes suggestions for narrowing the gap between the supply and demand of NEV charging infrastructure and the sustainable development of the NEV industry.

A KRR-UKF robust state estimation method for distribution networks

State estimation is an integral component of energy management systems. Employing a state estimation methodology that is both accurate and resilient is essential for facilitating informed decision-making processes. However, the complex scenarios (unknown noise, low data redundancy, and reconfiguration) of the distribution network pose new challenges for state estimation. In the context of this study, we introduce a state estimation technique known as the kernel ridge regression and unscented Kalman filter. In normal conditions, the non-linear correlation among data and unknown noise increases the difficulty of modeling the distribution network. Thence, kernel ridge regression is developed to map the data into high-dimensional space that transforms the non-linear problem into linear formulations base on the data rather the complicate grid model, which improves model generalization performance and filters out unknown noises. In addition, with the unique prediction correction mechanism of the Kalman method, the kernel ridge regression-mapped state value can be revised by the measurement, which further enhances model accuracy and robustness. During abnormal operating conditions and taking into account the presence of faulty data within the measurement system, we initiate the use of a long short-term memory network and combined convolutional neural network (CNN) model, referred to as the ATT-CNN-GRU. This model is utilized for the prediction of pseudo-measurements. Subsequently, we use an outlier detection method known as ordering points to identify the clustering structure to effectively identify and substitute erroneous data points. Cases on the IEEE-33 bus system and 109-bus system from a city in China show that the method has superior accuracy and robustness.

Multi-point deformation monitoring model of concrete arch dam based on MVMD and 3D-CNN

It is difficult for single measuring point model to synchronously characterize the spatial correlation of deformation sequence of concrete dam and the similarity of displacements of different measuring points. Considering the limitation of conventional model in dealing with spatial-temporal data structure, In this study, a deep learning technique for multi-point and multi-output deformation prediction of concrete dams is proposed, which is based on Multivariate Variational Mode Decomposition (MVMD) and three-dimensional convolutional neural network (3D-CNN) combined model. Firstly, the multi-point sequence is decomposed into several relatively stable sub-sequences with different frequency scales by MVMD, which reduces the complexity of the multi-point displacement sequence. Then, In order to enhance the model 's ability to fit the time and spatial features of the sequence, the 3D-CNN model is used to predict the decomposed components and reconstruct the displacement data of the measuring points. This study focuses on the principle, process and calculation steps of the proposed model, and verifies the proposed model through the deformation data of three typical measuring points. The test results show that the prediction model proposed in this paper is an effective method for dam deformation prediction.

Meteorological influences on PM2.5 variation in China using a hybrid model of machine learning and the Kolmogorov-Zurbenko filter

Quantitatively evaluating the contribution of meteorology and emissions to PM2.5 reduction during the Three-year Action Plan to Fight Air Pollution (2018–2020) is important to guide future emission control measures in China. In this study, a Kolmogorov-Zurbenko (KZ) filter and an artificial neural network (ANN) combined model was used to separate the meteorology-influenced PM2.5 concentration by considering seven meteorological factors including boundary layer height, surface pressure, surface net solar radiation, total precipitation, atmospheric temperature, wind speed and wind direction in China. To compensate for the black-box nature of the ANN, the Shapley additive explanation (SHAP) value was introduced to explain the effect of each meteorological factor on long-term PM2.5 concentration. Results showed that PM2.5 concentration decreased by 14% from 2018 to 2020 in China with annual average concentrations of 37.19, 35.28, and 31.94 μg/m3, respectively. Although the decline in PM2.5 was mainly due to emission reductions (58% for all of China), the contribution of prevailing meteorological conditions should not be ignored, especially in some regions such as Northeast China, which had the highest contribution of meteorology (58%) on PM2.5 reduction. In general, the contributions of meteorological conditions to the long-term trend of PM2.5 varied greatly by region and season. SHAP analysis indicated that temperature, surface pressure and boundary layer height were the dominant meteorological factors affecting long-term PM2.5 with contributions of 65%, 11% and 6% on average. We also found that regions with larger average PM2.5 concentration were more affected by emission changes than by meteorology. Therefore, we recommend the continuation of emission control measures in heavily polluted regions such as North and Central China.

CO2 emission forecasting based on nonlinear grey Bernoulli and BP neural network combined model

CO2 emission forecasting based on nonlinear grey Bernoulli and BP neural network combined model

Study of Improved Grey BP (Back Propagation) Neural Network Combination Model for Predicting Deformation in Foundation Pits

Deep excavation engineering is a comprehensive discipline that involves multiple fields such as engineering geology, hydrogeology, and foundation engineering. With the improvement of the utilization rate of underground space, the demand for the construction of large-scale underground structural engineering is growing, making the excavation of underground soil become increasingly frequent, which also brings about the safety problems of deep foundation pit engineering and the surrounding environment. Prediction of foundation pit deformation is an important research direction with diverse historical developments, but it is also facing a series of difficulties and challenges. In order to solve these problems, this article proposes an improvement plan, establishes a prediction model based on the combination model of grey BP (back propagation) neural network, and verifies its effectiveness through experiments. The results show that the average error of the new model’s prediction of horizontal deformation is about 0.31, which is about 32% lower than the traditional model’s prediction error. The difference between the vertical deformation prediction and actual monitoring results is also controlled. The vertical deformation predicted by wavelet transform is 7% to 9% larger than the actual monitoring results, meeting the prediction requirements. Finally, this article explores the research on the prediction of foundation pit deformation in deep excavation engineering, An improved grey BP neural network combination model was proposed and its effectiveness was verified through experiments. This article has important reference value for the study of deformation prediction in deep excavation engineering.

Mammography and ultrasound based dual modality classification of breast cancer using a hybrid deep learning approach

Traditional methods of diagnosing breast cancer (BC) suffer from human errors, are less accurate, and consume time. A computer-aided detection (CAD) system can overcome the above-stated limitations and help radiologists with accurate decision-making. However, the existing studies using single imaging modalities have shown limited clinical use due to its low diagnostic accuracy and reliability when compared to multimodal system. Thus, we aim to develop a hybrid deep learning bimodal CAD algorithm for the classification of breast lesions using mammogram and ultrasound imaging modalities combined. A combined convolutional neural network (CNN) and long-short term memory (LSTM) model is implemented using images from both mammogram and ultrasound modalities to improve the early diagnosis of BC. A new real-time dataset consisting of 43 mammogram images and 43 ultrasound images collected from 31 patients is used in this work. Further, each group consists of 25 benign and 18 malignant images. The number of images is increased to 1032 (516 for each modality) using different data augmentation techniques. The proposed bimodal CAD algorithm achieves a classification accuracy of 99.35% and the area under the receiver operating characteristic curve (AUC) of 0.99 over the traditional unimodal CAD systems, which attain the classification accuracy of 97.16% and 98.84% using mammogram and ultrasound, respectively. The proposed bimodal CAD algorithm using combined mammogram and ultrasound outperforms the traditional unimodal CAD systems. The bimodal CAD algorithm can avoid unnecessary biopsies and encourage its clinical application.

Application of MLR, BP and PCA-BP Neural Network for Predicting FeO in Bottom-Blowing O2-CaO Converter

In order to accurately predict the FeO content of slag in the bottom-blowing O2-CaO process of the dephosphorization converter, multiple linear regression model, backpropagation (BP) neural network model and principal component analysis–backpropagation (PCA-BP) combined with neural network model were established to predict the FeO content of slag. It was found that the PCA-BP combined neural network model has the highest prediction accuracy by using principal component analysis to reduce the dimension of influencing factors of FeO content in slag and eliminate the correlation between input variables. The average absolute error is 1.178%, which is 0.78% lower than that of multiple linear regression model and 0.453% lower than that of multiple linear regression model. When the prediction error range is 3.0%, the prediction hit rate of the model is 96%, and when the prediction error range is 2.0%, the prediction hit rate of the model is 78%. The prediction model has important reference value for actual production.

Probabilistic Forecasting of Residential Energy Consumption Based on SWT-QRTCN-ADSC-NLSTM Model

Residential electricity consumption forecasting plays a crucial role in the rational allocation of resources reducing energy waste and enhancing the grid-connected operation of power systems. Probabilistic forecasting can provide more comprehensive information for the decision-making and dispatching process by quantifying the uncertainty of electricity load. In this study, we propose a method based on stationary wavelet transform (SWT), quantile regression (QR), Bidirectional nested long short-term memory (BiNLSTM), and Depthwise separable convolution (DSC) combined with attention mechanism for electricity consumption probability prediction methods. First, the data sequence is decomposed using SWT to reduce the complexity of the sequence; then, the combined neural network model with attention is used to obtain the prediction values under different quantile conditions. Finally, the probability density curve of electricity consumption is obtained by combining kernel density estimation (KDE). The model was tested using historical demand-side data from five UK households to achieve energy consumption predictions 5 min in advance. It is demonstrated that the model can achieve both reliable probabilistic prediction and accurate deterministic prediction.

Based on BP neural network glass cultural relics chemical category and composition prediction model construction

The classification and prediction of chemical composition of glass cultural relics plays an important role in the study of cultural relics, and this paper obtains the chemical composition survey data of 58 groups of glass cultural relics, obtains the statistical law of the chemical composition content of glass cultural relics, and then builds a neural network prediction model of the chemical composition of glass cultural relics, and uses the chemical composition content of glass cultural relics as the training set and test set for building the prediction model. In this paper, the data is preprocessed and fed into the training set, so that the BP neural network model continues to learn, adjust the training parameters, and finally obtain the optimal prediction model. Comparing the predicted value with the real value shows that the BP neural network model has good accuracy. This experimental result verifies that the BP neural network combination model can effectively predict the category of glass cultural relics and predict the chemical content before and after differentiation, which provides an effective way for relevant departments in China to study the value of cultural relics.

Fusion deep learning approach combining diffuse optical tomography and ultrasound for improving breast cancer classification.

Diffuse optical tomography (DOT) is a promising technique that provides functional information related to tumor angiogenesis. However, reconstructing the DOT function map of a breast lesion is an ill-posed and underdetermined inverse process. A co-registered ultrasound (US) system that provides structural information about the breast lesion can improve the localization and accuracy of DOT reconstruction. Additionally, the well-known US characteristics of benign and malignant breast lesions can further improve cancer diagnosis based on DOT alone. Inspired by a fusion model deep learning approach, we combined US features extracted by a modified VGG-11 network with images reconstructed from a DOT deep learning auto-encoder-based model to form a new neural network for breast cancer diagnosis. The combined neural network model was trained with simulation data and fine-tuned with clinical data: it achieved an AUC of 0.931 (95% CI: 0.919-0.943), superior to those achieved using US images alone (0.860) or DOT images alone (0.842).

Prediction of buffeting responses of the thin plate under joint action of wave and wind using LSTM and transfer learning

To predicate the buffeting responses of long-span sea-crossing bridges under joint action of wave and wind, a thin plate was taken as the example, and a piston-type wave maker was adopted to simulate the second order Stokes wave in a computational domain. Through the large eddy simulation (LES) and dynamic mesh methods, time-histories of wind field and buffeting response of the thin plate under joint action of wave and wind were obtained. Then, the buffeting responses of the thin plate were predicted using the conventional long short-term memory (LSTM), convolutional LSTM (Conv LSTM) and LSTM with attention mechanism (LSTM-AM) neural network models, respectively. Furthermore, the transfer learning (TL) method was introduced, the method to solve a large amount of source task data was established, and a combined neural network model of TL-Conv LSTM-AM was proposed to predict the buffeting responses of the thin plate. The results show that the prediction accuracies of Conv LSTM and LSTM-AM models are both higher than the conventional LSTM model based on the evaluation indexes of determination coefficient (R2) and root mean square error (RMSE). The prediction accuracies of the buffeting responses can be improved significantly when the above LSTM neural network models are combined with the TL method. Also, there are still some deviations between the predicated time-histories and target ones. For the proposed combined model of TL-Conv LSTM-AM, the values of R2 in three directions of the buffering responses are all larger than 0.996 and 0.983 in the high and low turbulent flow fields, respectively. Therefore, the proposed TL-Conv LSTM-AM combined model has the highest prediction accuracy among them. Similar prediction accuracy can be achieved for longer time duration of the testing data set using the proposed combined model of TL-Conv LSTM-AM. The above phenomenon proves that the effectiveness of this proposed combined model in predicting the buffeting responses of the thin plate under joint action of wave and wind.

Reactor Temperature Prediction Method Based on CPSO-RBF-BP Neural Network

A neural network model based on a chaotic particle swarm optimization (CPSO) radial basis function-back propagation (RBF-BP) neural network was suggested to improve the accuracy of reactor temperature prediction. The training efficiency of the RBF-BP neural network is influenced to some degree by the large randomness of the initial weight and threshold. To address the impact of initial weight and threshold uncertainty on the training efficiency of the RBF-BP combined neural network, this paper proposes using a chaotic particle swarm optimization algorithm to correct the RBF-BP neural network’s initial weight and threshold, as well as to optimize the RBF-BP neural network to speed up the algorithm and improve prediction accuracy. The measured temperature of the reactor acquired by on-site enterprises was confirmed and compared to the predicted results of the BP, RBF-BP, and PSO-RBF-BP neural network models. Finally, Matlab simulation tests were performed, and the experimental data revealed that the CPSO-RBF-BP combined neural network model suggested in this paper had a root-mean-square error of 17.3%, an average absolute error of 11.4%, and a fitting value of 99.791%. Prediction accuracy and efficiency were superior to those of the BP, RBF-BP, and PSO-RBF-BP models. The suggested model’s validity and feasibility were established. The study findings may provide some reference values for the reactor’s temperature prediction.