Number Of Nodes In Hidden Layer Research Articles

Firefly Interval Selection Combined With Extreme Learning Machine for Spectral Quantification of Complex Samples

ABSTRACTFirefly algorithm (FA) combined with extreme learning machine (ELM) is developed for spectral interval selection and quantitative analysis of complex samples. The method firstly segments the spectra into a certain number of intervals. Vectors with 1 and 0, which represent the interval selected or not, are used as the inputs of the FA. The RMSEP value predicted by ELM model is used as the fitness function of the FA. The activation function and number of hidden layer nodes of ELM, number of spectral intervals, population number, environmental absorbance, and constant of FA are optimized. The predictive performance of FA‐ELM is compared with full‐spectrum PLS, ELM, genetic algorithm‐ELM (GA‐ELM), and particle swarm optimization‐ELM (PSO‐ELM) by one ultraviolet (UV) spectrum dataset of gasoil and three near‐infrared (NIR) spectral datasets of corn, wheat, and tablet samples. The results show that FA‐ELM has a better performance compared with its competitors in predicting monoaromatics, water, wheat kernel texture, and active pharmaceutical ingredients (APIs) in gasoil, corn, wheat, and tablet samples.

Performance prediction model for desalination plants using modified grey wolf optimizer based artificial neural network approach

Desalination represents an effective method for alleviating water scarcity, applying algorithmic techniques to predict the performance of reverse osmosis (RO) desalination plants, Modified Grey Wolf Optimizer (MGWO) based Artificial Neural Networks (ANN) can predict the performance of membrane distillation (MD) equipment. Four experimental inputs are selected: feed salt concentration(35–140 g/h), feed flow rate(400–600 L/h), evaporator inlet temperature (60–80 ℃), and condenser inlet temperature (20–30 ℃). The permeate flux (L/h m2) is selected as the experimental output. Ten prediction models were proposed and compared with the existing models (ANN, WOA-ANN, and GWO-ANN). The results showed that the MGWO-ANN model-5 showed the best regression results: R2 = 99.3 %, mean square error (MSE)= 0.004. This model outperformed the existing ANN model (R2 =98.8 %, MSE=0.060), WOA-ANN model (R2 =99.1 %, MSE=0.005) and GWO-ANN model (R2 =98.9 %, MSE=0.007). Model-5 has a single hidden layer (H=1), 13 hidden nodes (n = 13), 10 search agents (SA=10), and 75 %− 20 %− 05 % dataset division. Its residual error is within acceptable limits (spanning −0.1 to 0.2). Optimizing the number of hidden layer nodes (n) and the number of search agents (SA) can improve the training efficiency and prediction accuracy of the model, and the MGWO-ANN model is capable of more accurately predicting the performance of desalination plants.

Algorithm for community security risk assessment and influencing factors analysis by back propagation neural network

This paper aims to accurately assess and effectively manage various security risks in the community and overcome the challenges faced by traditional models in handling large amounts of features and high-dimensional data. Hence, this paper utilizes the back propagation neural network (BPNN) to optimize the security risk assessment model. A key challenge of researching community security risk assessment lies in accurately identifying and predicting a range of potential security threats. These threats may encompass natural disasters, public health crises, accidents, and social security issues. The intricate interplay of these risk factors, combined with the dynamic nature of community environments, presents difficulties for traditional risk assessment methodologies to address effectively. Initially, this paper delves into the factors influencing safety incidents within communities and establishes a comprehensive system of safety risk assessment indicators. Leveraging the adaptable and generalizable nature of the BPNN model, the paper proceeds to optimize the BPNN model, enhancing the security risk assessment model through this optimization. Subsequent comparison experiments with traditional models validate the rationality and effectiveness of the proposed model, with hidden layer nodes set at various levels like 10, 15, 20, 25, 30, and 35. These traditional models include Convolutional Neural Network (CNN), Long Short-Term Memory Network (LSTM), Bidirectional Encoder Representations from Transformers (BERT), Generative Pre-trained Transformer (GPT), and eXtreme Gradient Boosting (XGBOOST). Experimental findings demonstrate that with 20 hidden layer nodes, the optimized model achieves a remarkable final recognition accuracy of 99.1 %. Moreover, the optimized model exhibits significantly lower final function loss compared to models with different node numbers. Increasing the number of hidden layer nodes may diminish the optimized model's fit and accuracy. Comparison with traditional models reveals that the average accuracy of the optimized model in community risk identification reaches 98.5 %, with a maximum accuracy of 99.6 %. This marks an improvement of 9%–11 % in recognition accuracy across various risk factors compared to traditional models. Regarding system response time and resource utilization, the optimized model exhibits a response time ranging from 100 ms to 120 ms and consistently lower resource utilization rates across all scenarios, underscoring its efficiency in community security risk assessment. In conclusion, this experiment sheds light on the underlying mechanisms and patterns of community safety risk formation, offering novel perspectives and methodologies for researching community safety risk assessment. The paper concludes by presenting recommendations and strategies for addressing community safety risks based on experimental analysis.

Performance Prediction and Optimization of Single-Piston Free Piston Expander-Linear Generator Based on Machine Learning and Genetic Algorithm

This paper proposed a single-piston free piston expander-linear generator (SFPE-LG) prototype applied to organic Rankine cycle systems. Two valve timing control strategies, namely, time control strategy (TCS) and position control strategy (PCS), were developed. Based on the experimental data, a back propagation neural network (BPNN) prediction model was established. The effects of structural parameters such as neural network layers, transfer function, training function, hidden layer nodes, and learning rate on the prediction accuracy of this BPNN model were discussed. The training and prediction accuracy of the BPNN model was verified using 5-fold cross-validation and Wilcoxon signed-rank test. Moreover, the BPNN model was integrated with a genetic algorithm to predict and optimize the maximum output power of the SFPE-LG. The results showed that the BPNN model used to predict the motion characteristics and output performance of the SFPE-LG exhibits strong learning ability and high prediction accuracy. Notably, the prediction accuracy of the BPNN model is significantly higher under the PCS compared to TCS. The effect of hidden layer nodes on mean square error (MSE) and correlation coefficient (R) is greater than that of the learning rate. When the number of hidden layer nodes exceeds 30, the BPNN model consistently achieves low MSE and high R. The optimization results showed that the SFPE-LG can obtain a maximum output power of 141.69 W under the TCS, when the working parameters are inlet pressure of 0.7 MPa, intake duration of 35 ms, load resistance of 67 Ω, and expansion duration of 104 ms, respectively.

A cooperative stochastic configuration network based on differential evolutionary sparrow search algorithm for prediction

Stochastic configuration network (SCN) is a powerful prediction model whose performance is significantly influenced by the configuration of the network parameters. To improve the prediction accuracy of the network, a cooperative stochastic configuration network (CSCN) based on a novel differential evolutionary sparrow search algorithm (DESSA), termed as DESSA-CSCN, is proposed. In the CSCN, the number of hidden layer nodes is adaptively adjusted according to the number of iterations, and the parameters of hidden nodes are cooperatively optimized by using a population-based metaheuristic algorithm. A sparse matrix is introduced to mitigate parameter overfitting caused by the increased number of hidden layer nodes. During parameter optimization, the fitness function is constructed by using the supervision mechanism of the SCN, and the DESSA is utilized as the metaheuristic algorithm to update the weights and biases. In order to verify the effectiveness of the DESSA-CSCN, several simulation experiments have been conducted. The performance of the DESSA is evaluated by the CEC2017 test suit, and the simulation results show that the DESSA exhibits better convergence accuracy and can jump out of local optima more effectively than other algorithms. The performance of the DESSA-CSCN is evaluated by 4 datasets from KEEL, and the simulation results indicate that the DESSA-CSCN achieves better prediction accuracy and faster prediction speed than other models.

Creation of Deep Learning Scenarios in the Network Teaching of Physical Education Technical Courses

The network teaching evaluation of sports professional technical courses has positive significance for the sustainable development of education. And how to establish an effective evaluation model is the key part. The research introduces the creation of in-depth learning scenarios (LS) into the network teaching of sports professional technical courses, and then constructs a new network teaching mode of sports professional technical courses. The Particle Swarm Optimization algorithm - Attention- Long Short-Term Memory network (PSO-Attention-LSTM) Chinese Emotion Classification Model (ECM) is constructed to classify the online evaluation text to realize the evaluation of online teaching. This model combines the improved PSO, Attention and LSTM classification models. The optimal number of hidden layer nodes for LSTM model is about 100, and the optimal data size for batch processing is 25. The overall error rate of online teaching teachers of male and female sports professional skill courses is 10.1%, and the overall error rate of online teaching teachers of general and advanced sports professional skill courses is 12.8%. The application effect of the creation of in-depth learning scene in the network teaching of physical education technical courses is shown. When the classification threshold is 0.6 and 0.8 respectively, the AUC of PSO-Attention-LSTM Chinese ECM is 0.821 and 0.809 respectively. The research institute has put forward that the online teaching platform of sports professional technical courses has extremely high practical application value and has been unanimously praised by network users.

An Improved Extreme Learning Machine (ELM) Algorithm for Intent Recognition of Transfemoral Amputees With Powered Knee Prosthesis.

To overcome the challenges posed by the complex structure and large parameter requirements of existing classification models, the authors propose an improved extreme learning machine (ELM) classifier for human locomotion intent recognition in this study, resulting in enhanced classification accuracy. The structure of the ELM algorithm is enhanced using the logistic regression (LR) algorithm, significantly reducing the number of hidden layer nodes. Hence, this algorithm can be adopted for real-time human locomotion intent recognition on portable devices with only 234 parameters to store. Additionally, a hybrid grey wolf optimization and slime mould algorithm (GWO-SMA) is proposed to optimize the hidden layer bias of the improved ELM classifier. Numerical results demonstrate that the proposed model successfully recognizes nine daily motion modes including low-, mid-, and fast-speed level ground walking, ramp ascent/descent, sit/stand, and stair ascent/descent. Specifically, it achieves 96.75% accuracy with 5-fold cross-validation while maintaining a real-time prediction time of only 2 ms. These promising findings highlight the potential of onboard real-time recognition of continuous locomotion modes based on our model for the high-level control of powered knee prostheses.

การพยากรณ์ยอดขายและความต้องการสินค้าในอุตสาหกรรมผลิตอุปกรณ์ไฟฟ้า กรณีศึกษา บริษัทผลิตอุปกรณ์ไฟฟ้าแรงสูง

The purpose of this study is to find the product demand forecasting model from the artificial neural network method in a case study of a high-voltage equipment manufacturing company. And reduce fines for late delivery of goods by using the Matlab program. The forecasting model is established by using time series neural network method, and the nonlinear autoregressive neural network (NARX) function model is established by combining with external input data. Then the forecast deviation is the average deviation percentage error (MAPE) uses the least dynamics model. By using data that collects product sales from 2019 – 2022 total 48 set of data are divided into 2 data sets used in the study. by the first dataset is learning 36 datasets and the second dataset is learning 12 datasets. The model of this study was defined with the number of hidden layer nodes as 2, 6 and 10. The time delay was 2 and 3. Forecasting models are combined into 6 models. The best model result of each item from NARX forecast through the learning process Levenberg-Marquardt (trainlm), including models 1, 4, 5, 4, 3, 3, and 2 sequential The results show the least average absolute percentage error at 8.32, 5.85, 2.61, 3.39, 2.91, 2.19 and 7.76 and the results of the forecasting model were analyzed to reduce fines from selling all 7 items Able to reduce penalties for late delivery of 4 items showing the percentage of difference in cost reduction equal to 0.28% 0.24% 0.32% and 0.22%. It can be concluded that the prediction of the neural network inverted neural network model It can be used to forecast product sales and product demand as well. And reduce the cost of late delivery fines for better results.

A modeling approach for the absorber of absorption heat pump systems

The study on the data-and-knowledge-driven modeling methods is very important for the digital twinning and the optimal control of absorption heat pump systems. For the purpose of obtaining an absorber model with high accuracy, low computational effort and less time-consuming for the absorption heat pump systems, a modeling method is proposed by combining the Simple Particle Swarm Optimization (SPSO) and the Wavelet Neural Network (WNN). The input-output structure of the model is presented by analyzing the operational principles of the absorber. In view of the complexity of the heat transfer and mass transfer processes inside the absorber, a single hidden layer WNN is employed as the internal structure of the model. The number of hidden layer nodes in the network model is determined by the trial-and-error method using the root mean square error and coefficient of determination. To overcome the drawback of WNN which can easily fall into the local optima during the error backpropagation process and improve the prediction accuracy of the model, the Morlet wavelet basis function and a SPSO algorithm are exploited respectively to replace the S-function of Neural Network and determine the optimal parameter values of the neural network (e.g. the weight, scaling factor, and translation factor, etc.). Experimental simulation results show the root mean square errors of the proposed model are 1.59 % and 2.21 % for the two outputs, respectively, which are smaller than the conventional WNN model. The proposed modeling approach effectively solves the problems of the BP neural network on easy falling into the local optimum and the slow convergence, which will be widely used in the process of the optimized operation control for heat pump systems.

Effective fault module localization in substation critical equipment: an improved ant colony optimization and back propagation neural network approach

AbstractThe rapid development of substations has increased the demand for accurate and fast fault prediction systems. In order to achieve rapid localization and autonomous decision‐making of fault modules and types in substations, the article proposes a fault autonomous localization algorithm based on improved ant colony optimization (IACO) and back propagation neural network (BPNN). The fault data of the substation secondary equipment for training and testing the BPNN model is based on the actual operating equipment of the substation, which can significantly improve the reliability of the model results. In addition, the IACO is used to globally optimize the weights and thresholds of BPNN, and the number of hidden layer nodes in BPNN was analyzed to further improve the accuracy of the established fault prediction algorithm. The test results show that the fault prediction accuracy of the BPNN model optimized by IACO is 93.67%, which is significantly improved compared to the traditional BPNN and BPNN with ant colony optimization method (with an accuracy of 82.98% and 91.04%). The above results effectively demonstrate the high accuracy and effectiveness of the established prediction algorithm in processing data and locating faults, which can improve the maintenance and operational efficiency of substations.

The Successful Prediction Of Waterflooding Using A Feed Forward Algorithm

Waterflooding is one of the most frequently used EOR methods to increase oil recovery because it can increase 30% -60% of total production. It is necessary to apply a production system performance prediction approach to minimize uncertainty in increasing production figures, such as analytical methods and numerical methods. Artificial Intelligence in the world of oil and gas is not a new thing, but it has often been used in various fields such as exploration, drilling, production, and reservoirs. So this is the basis for the prediction of the success of waterflooding research carried out. The purpose of this research was to predict the success rate of waterflooding using an Artificial Neural Network (ANN). The method used in this study is the simulation research method using CMG Imex for reservoir simulation modeling, running CMG CMOST for 500 sensitivity data with the input of seven parameters of compressibility, horizontal permeability, vertical permeability, pressure injection, injection rate, thickness, oil saturation, and the output is recovery factor using Artificial Neural Network (ANN) with a ratio of 70% of the RF calculation model results for training and 30% model results for testing. In order to get optimal prediction results, trial, and error were carried out on the number of hidden layer nodes, so that optimal and stable hidden layer nodes were obtained at node 10 with RMSE values of 0.339035 for training and 0.442663 for testing and MAPE for training 1.15% and 1.62% for testing. The statistical analysis value is 0.906139 for training and 0.899525 for testing data. It can be concluded from this study that the use of ANN in predictions using ten hidden layer nodes proved to be very good and successful, and predictions in this study were classified as High Accurate Prediction.

Towards an Optimal KELM Using the PSO-BOA Optimization Strategy with Applications in Data Classification.

The features of the kernel extreme learning machine-efficient processing, improved performance, and less human parameter setting-have allowed it to be effectively used to batch multi-label classification tasks. These classic classification algorithms must at present contend with accuracy and space-time issues as a result of the vast and quick, multi-label, and concept drift features of the developing data streams in the practical application sector. The KELM training procedure still has a difficulty in that it has to be repeated numerous times independently in order to maximize the model's generalization performance or the number of nodes in the hidden layer. In this paper, a kernel extreme learning machine multi-label data classification method based on the butterfly algorithm optimized by particle swarm optimization is proposed. The proposed algorithm, which fully accounts for the optimization of the model generalization ability and the number of hidden layer nodes, can train multiple KELM hidden layer networks at once while maintaining the algorithm's current time complexity and avoiding a significant number of repeated calculations. The simulation results demonstrate that, in comparison to the PSO-KELM, BBA-KELM, and BOA-KELM algorithms, the PSOBOA-KELM algorithm proposed in this paper can more effectively search the kernel extreme learning machine parameters and more effectively balance the global and local performance, resulting in a KELM prediction model with a higher prediction accuracy.

A New Method for Calculating Prediction Parameters of Surface Deformation in the Mining Area

The accurate calculation of mining-induced surface deformation has important guiding significance for efficient and safe production in mining areas. The probability integral method (PIM) is a main prediction method in China, and the selection of its parameters is directly related to the prediction accuracy of surface deformation in mining areas. To overcome shortcomings of PIM and other methods, this paper proposed a prediction model of the parameters of PIM combining a multiple regression model and an extreme learning machine. In this paper, the Huainan mining area was selected as the research object, the influence factors of PIM parameters were analyzed and the accuracy of the model was verified. The influence of the number of hidden layer nodes, the selection of activation function and the proportion of training set and test set in the model were analyzed. The conclusions suggest that the PIM parameters calculated in this paper could be used to predict mining subsidence and obtain surface movement and deformation data. The research results provide an effective method for the selection of surface deformation prediction parameters of new working faces or faces lacking measured data.

Gearbox Fault Diagnosis Based on Optimized Stacked Denoising Auto Encoder and Kernel Extreme Learning Machine

The gearbox is one of the key components of many large mechanical transmission devices. Due to the complex working environment, the vibration signal stability of the gear box is poor, the fault feature extraction is difficult, and the fault diagnosis accuracy makes it difficult to meet the expected requirements. To solve this problem, this paper proposes a gearbox fault diagnosis method based on an optimized stacked denoising auto encoder (SDAE) and kernel extreme learning machine (KELM). Firstly, the particle swarm optimization algorithm in adaptive weight (SAPSO) was adopted to optimize the SDAE network structure, and the number of hidden layer nodes, learning rate, noise addition ratio and iteration times were adaptively obtained to make SDAE obtain the best network structure. Then, the best SDAE network structure was used to extract the deep feature information of weak faults in the original signal. Finally, the extracted fault features are fed into KELM for fault classification. Experimental results show that the classification accuracy of the proposed method can reach 97.2% under the condition of low signal-to-noise ratio, which shows the effectiveness and robustness of the proposed method compared with other diagnostic methods.

Prediction of the corrosion depth of oil well cement corroded by carbon dioxide using GA-BP neural network

A corrosion prediction model was established based on the genetic algorithm (GA) and back propagation (BP) neural network to predict the long-term corrosion changes of oil well cement, Considering that the cement sheath is susceptible to corrosion and its corrosion degree is not easy to observe in acid gas wells and geological storage wells containing carbon dioxide (CO2). The initial weights and thresholds of the neural network were optimized by GA. The number of hidden layer nodes was selected by error verification, and the network was trained with an improved algorithm. The sample data was regression processed based on the empirical formula and was used in the network training. The simulation results shows that: The improved GA-BP network model (3–5-6–1) has a higher prediction accuracy with faster convergence and better fitting effect compared with the traditional BP neural network and the regression model (REG) in long-term prediction of corrosion depth in oil well cement. The regression coefficient (R2) of the prediction model is 0.9913, and the mean square error (MSE) of test samples is 0.0026. The modeling idea proposed in this paper can be applied to improve the accuracy of prediction models in predicting the corrosion of oil well cement.

State-of-health estimation of Lithium-ion battery based on back-propagation neural network with adaptive hidden layer

The reliability and safety of lithium-ion batteries (LIBs) are key issues in battery applications. Accurate prediction of the state-of-health (SOH) of LIBs can reduce or even avoid battery-related accidents. In this paper, a new back-propagation neural network (BPNN) is proposed to predict the SOH of LIBs. The BPNN uses as input the LIB voltage, current and temperature, as well as the charging time, since it is strongly correlated with the SOH. The number of hidden layer nodes is adaptively set based on the training data in order to improve the generalization capability of the BPNN. The effectiveness and robustness of the proposed scheme is verified using four distinct battery datasets and different training data. Experimental results show that the new BPNN is able to accurately predict the SOH of LIBs, revealing superiority when compared to other alternatives.

STWD-SFNN: Sequential three-way decisions with a single hidden layer feedforward neural network

The three-way decisions strategy has been employed to construct network topology in a single hidden layer feedforward neural network (SFNN). However, this model has a general performance, and does not consider the process costs, since it has fixed threshold parameters. Inspired by the sequential three-way decisions (STWD), this paper proposes STWD with an SFNN (STWD-SFNN) to enhance the performance of networks on structured datasets. STWD-SFNN adopts multi-granularity levels to dynamically learn the number of hidden layer nodes from coarse to fine, and set the sequential threshold parameters. Specifically, at the coarse granular level, STWD-SFNN handles easy-to-classify instances by applying strict threshold conditions, and with the increasing number of hidden layer nodes at the fine granular level, STWD-SFNN focuses more on disposing of the difficult-to-classify instances by applying loose threshold conditions, thereby realizing the classification of instances. Moreover, STWD-SFNN considers and reports the process cost produced from each granular level. The experimental results verify that STWD-SFNN has a more compact network on structured datasets than other SFNN models, and has better generalization performance than the competitive models. All models and datasets can be downloaded from https://github.com/wuc567/Machine-learning/tree/main/STWD-SFNN.

A novel correlation Gaussian process regression-based extreme learning machine.

An obvious defect of extreme learning machine (ELM) is that its prediction performance is sensitive to the random initialization of input-layer weights and hidden-layer biases. To make ELM insensitive to random initialization, GPRELM adopts the simple an effective strategy of integrating Gaussian process regression into ELM. However, there is a serious overfitting problem in kernel-based GPRELM (kGPRELM). In this paper, we investigate the theoretical reasons for the overfitting of kGPRELM and further propose a correlation-based GPRELM (cGPRELM), which uses a correlation coefficient to measure the similarity between two different hidden-layer output vectors. cGPRELM reduces the likelihood that the covariance matrix becomes an identity matrix when the number of hidden-layer nodes is increased, effectively controlling overfitting. Furthermore, cGPRELM works well for improper initialization intervals where ELM and kGPRELM fail to provide good predictions. The experimental results on real classification and regression data sets demonstrate the feasibility and superiority of cGPRELM, as it not only achieves better generalization performance but also has a lower computational complexity.

A Back Propagation Neural Network-Based Radiometric Correction Method (BPNNRCM) for UAV Multispectral Image

Radiometric correction is one of the most important preprocessing parts of unmanned aerial vehicle (UAV) multispectral remote sensing data analysis and application. In this article, a back propagation (BP) neural network-based radiometric correction method (BPNNRCM) considering optimal parameters was proposed. First, we used different UAV multispectral sensors (K6 equipped on the DJI M600, D-MSPC2000 equipped on the FEIMA D2000) to collect training, validation, testing and cross-validation data. Second, the radiometric correction results of BP neural network with different input variables and hidden layer node number were compared to select the best combination of input parameters and hidden layer node number. Finally, the radiometric correction accuracy and robustness of BP neural network considering the optimal parameters were verified. When the number of nodes in the input layer was five (digital number, UAV sensor height, wavelength, solar altitude angle, and temperature) and the number of nodes in the hidden layer was eight, the BP neural network had the best comprehensive performance in training time of train set and accuracy of validation/test set. In the aspect of accuracy and robustness, the absolute errors of test and cross-validation images' surface reflectance obtained by the BPNNRCM were all less than 0.054. The BPNNRCM had smaller average absolute error (0.0141), mean squared error (0.0003), mean absolute error (0.0141) and mean relative error (7.1%) comparing with empirical line method and radiative transfer model. In general, the research results of this article prove the feasibility and prospect of BPNNRCM for radiometric correction of UAV multispectral images.

An Intelligent Adequate-Fitting Prediction Method of Coastal Tunnel Rock Deformation Based on the Effective Rank of Hidden Layer

The intelligent prediction of surrounding rock deformation is of great significance for guiding the design and construction of tunnel projects in coastal areas. The deformation of tunnels in coastal areas is more complex than that of the ground, and the risk of encountering adverse geological conditions is greater. The traditional tunnel deformation prediction method contains the defects of a fixed model, a limited sample number, and it is easy to fall into underfitting and local overfitting. Therefore, the capacity of previous methods is limited by significant error, weak generalization, and poor intelligence. This paper proposes an adequate fitting prediction method for tunnel deformation based on the effective rank theory of the hidden layer nodes’ output matrix to analyze the surrounding rock and predict its deformation intelligently. Based on the traditional BPNN (back propagation neural network) algorithm, the number of hidden layer nodes is determined by the effective rank of the output matrix. Then, the approximation error and degree were adopted to reflect the approximation law of the BPNN to achieve the purpose of overfitting and underfitting control. An optimized BP neural network model for intelligently predicting tunnel deformation is constructed. Then, the optimized BPNN model is applied to a case study of a coastal tunnel in South China. Compared with the prediction method of LR (linear regression) and TS (time series), the results show that the prediction results of the optimized model are in good agreement with the measured values, with strong generalization ability and high intelligence. The proposed method is of guidance to other tunnels surrounding rock deformation prediction and engineering practice.