Number Of Neurons In Hidden Layer Research Articles

A low-cost and efficient electronic nose system for quantification of multiple indoor air contaminants utilizing HC and PLSR

A novel quantification technique is presented for electronic nose (E-nose), which is based on a double-step strategy combined with hierarchical classifier (HC) and partial least squares regression (PLSR). With the tree structure of HC, the complexity of classifier training process can be reduced in the case of unbalanced samples. For each level of the class hierarchy, the extreme learning machine-based artificial neural network (ELM-ANN) is applied for classification. In order to improve the classification performance of ELM-ANN, the multiple time-domain features are selected as training inputs, and a novel optimization method of the number of hidden layer neurons is given. To validate the effectiveness of the presented quantification technique, an E-nose system is designed to quantify the gases including six toxic gases (hydrogen sulfide, carbon monoxide, ammonia, toluene, formaldehyde, acetone) and three kinds of binary gas mixtures. This presented hierarchical classifier has demonstrated outstanding performance for the identification of target gases, such as the macro-averaged precision for unlabeled data is improved from 80% to 92% compared with non-hierarchical classifiers. Furthermore, an excellent performance of concentration estimation is obtained utilizing PLSR, where average values of the coefficient of determination for training and test samples are equal to 0.957 and 0.927, respectively. Overall, our work demonstrates that the proposed approach is applicable in E-nose-based odor quantification.

Evaluation of ice-seabed interaction mechanism in sand by using self-adaptive evolutionary extreme learning machine

Recent discovered oil and gases in the Arctic area have heightened the need for more attention to ice-seabed interaction during an ice scouring event. The seabed is gouged by these drifting icebergs in warmer months threatening the subsea pipelines transferring the hydrocarbons from offshore to onshore. The simulation of ice scouring seabed needs costly large deformation finite element analysis for the guaranteed operational integrity of the subsea pipelines. In this paper, a cost-effective alternative approach using the self-adaptive evolutionary extreme learning machine (SaE-ELM) algorithm was taken to model the ice-induced seabed scour. Initially, using parameters governing the ice gouging process, 17 SaE-ELM models were developed. Then, a comprehensive dataset was established and properly allocated for training and testing of the developed models. The optimal number of hidden layer neurons and the best activation function were opted for the SaE-ELM network. The premium SaE-ELM models and the most influencing inputs were recognized by conducting a sensitivity analysis. The vertical component of load showed a significant impact on the reaction forces, rather the soil depth and berm height possessed a striking effect for modeling the soil displacements. Ultimately, a set of the SaE-ELM-based equations were presented to estimate the subgouge soil parameters.

Research on Acoustic Emission Signal Denoising Based on Autoencoder

Acoustic emission signal denoising is the premise of acoustic emission signal identification. The traditional filter and wavelet analysis have the problems of relying on prior information and poor adaptability in acoustic emission signal denoising. Therefore, a noise denoising model of acoustic emission signal based on noise reduction autoencoder is proposed. By unsupervised learning training, the noise reduction autoencoder has more stable invariant characteristics, so that the error between the reconstructed signal and the original signal converges to a minimum, thus achieving the purpose of denoising. The denoising experiments are carried out on the basis of processing 3000 corrosive acoustic emission signal samples. The experimental results show that the model has better denoising effect when the number of hidden layer neurons is 400. The proposed DAE model has better performance and robustness than Donoho wavelet threshold denoising method.

An Effective Turkey Marble Classification System: Convolutional Neural Network with Genetic Algorithm -Wavelet Kernel - Extreme Learning Machine

Marble is one of the most popular decorative elements. Marble quality varies depending on its vein patterns and color, which are the two most important factors affecting marble quality and class. The manual classification of marbles is likely to lead to various mistakes due to different optical illusions. However, computer vision minimizes these mistakes thanks to artificial intelligence and machine learning. The present study proposes the Convolutional Neural Network- (CNN-) with genetic algorithm- (GA) Wavelet Kernel- (WK-) Extreme Learning Machine (ELM) (CNN–GA-WK-ELM) approach. Using CNN architectures such as AlexNet, VGG-19, SqueezeNet, and ResNet-50, the proposed approach obtained 4 different feature vectors from 10 different marble images. Later, Genetic Algorithm (GA) was used to optimize adjustable parameters, i.e. k, 1, and m, and hidden layer neuron number in Wavelet Kernel (WK) – Extreme Learning Machine (ELM) and to increase the performance of ELM. Finally, 4 different feature vector parameters were optimized and classified using the WK-ELM classifier. The proposed CNN–GA-WK-ELM yielded an accuracy rate of 98.20%, 96.40%, 96.20%, and 95.60% using AlexNet, SequeezeNet, VGG-19, and ResNet-50, respectively.

Network intrusion detection based on IE-DBN model

Existing network intrusion detection models suffer such problems as low detection accuracy and high false alarm rates in face of massive data traffic. Deep-learning models provide a solution as they can reduce the dimensionality of massive data, extract data features, and identify intrusions. However, the network structure and the number of hidden layer neurons of deep-learning models are determined by empirical or trial-and-error methods, which will affect the generalization ability and learning efficiency of the model. In the present work, a deep belief network model based on information entropy (IE-DBN model) is proposed for network intrusion detection. The model uses information gain (IG) to reduce the dimensionality of high-dimensional data features and remove redundant features. The information entropy is used to determine the number of hidden neurons in the DBN network and the network depth. The synthetic minority oversampling technique (SMOTE) algorithm is used to address the problem of data imbalance. Tests on the KDD CUP 99 intrusion detection data set have shown that the proposed IE-DBN model improved the convergence speed of the model and reduced the likelihood of overfitting. Compared with the conventional back propagation (BP) neural network and DBN network model, the IE-DBN model obtained a higher detection accuracy and a lower false alarm rate. Verification tests on other intrusion detection data sets showed that the proposed IE-DBN model had good generalization capacity.

Prediction of the Compressive Strength of Recycled Aggregate Concrete Based on Artificial Neural Network.

Recycled aggregate concrete (RAC), due to its high porosity and the residual cement and mortar on its surface, exhibits weaker strength than common concrete. To guarantee the safe use of RAC, a compressive strength prediction model based on artificial neural network (ANN) was built in this paper, which can be applied to predict the RAC compressive strength for 28 days. A data set containing 88 data points was obtained by relative tests with different mix proportion designs. The data set was used to develop an ANN, whose optimal structure was determined using the trial-and-error method by taking cement content (C), sand content (S), natural coarse aggregate content (NCA), recycled coarse aggregate content (RCA), water content (W), water–colloid ratio (WCR), sand content rate (SR), and replacement rate of recycled aggregate (RRCA) as input parameters. On the basis of different numbers of hidden layers, numbers of hidden layer neurons, and transfer functions, a total of 840 different back propagation neural network (BPNN) models were developed using MATLAB software, which were then sorted according to the correlation coefficient R2. In addition, the optimal BPNN structure was finally determined to be 8–12–8–1. For the training set, the correlation coefficient R2 = 0.97233 and RMSE = 2.01, and for the testing set, the correlation coefficient R2 = 0.96650 and RMSE = 2.42. The model prediction deviations of the two were both less than 15%, and the results show that the ANN achieved pretty accurate prediction on the compressive strength of RAC. Finally, a sensitivity analysis was carried out, through which the impact of the input parameters on the predicted compressive strength of the RAC was obtained.

Prediksi Penyakit Ginjal Kronis Menggunakan Hibrid Jaringan Saraf Tiruan Backpropagation dengan Particle Swarm Optimization

The number of Chronic Kidney Disease patient increased year by year while it doesn’t following by sufficient human resources and infrastructure needs the information of Chronic Kidney Disease patient prediction. Prediction of Chronic Kidney Disease patient is necessary to be done as an anticipation for preparing the better human resources and infrastructure that will effect to patient survival rate. In this study, backpropagation artificial neural network and particle swarm optimization combination used to predict the number of Chronic Kidney Disease patient. Artificial Neural Network has the ability in time series data prediction, such as the number of Chronic Kidney Disease year by year. But, backpropagation artificial neural network has a weakness in weight inisialization which taken unoptimally that could cause bad convergence speed. Particle swarm optimization will resolve the backpropagation artificial neural network weakness by weights optimization that will used in backpropagation artificial neural network. The Artificial Neural Network and Particle Swarm Optimization have several parameters, such as the number of hidden layer neuron, learning rate, and swarm. This research is using RSUD Banyumas Chronic Kidney Disease patient data in 2011 until 2020. Matlab R2019a used in this research as a software to predict chronic kidney disease patient data. The test results shows the prediction accuracy based on Mean Squared Error value is 0,0370 using 12-16-1 artificial neural network architecture, 0.005 learning rate, 1250 epochs and 50 swarms

An efficient and effective deep convolutional kernel pseudoinverse learner with multi-filter

The convolutional neural network is the most widely used deep neural network. However, it still has some disadvantages. First, the back-propagation method is usually used in the training of convolutional neural networks, but it has several inherent defects, such as the vanishing gradient problem and exploding gradient problem. Moreover, the training of a convolutional neural network often needs substantial computational resources and time. To solve the above problems, based on pseudoinverse learning (PIL), kernel pseudoinverse learning (KPIL) is proposed, showing improved performance. The number of hidden layer neurons of KPIL is equal to the number of input data and does not need to be set. KPIL uses the kernel method to calculate the output of hidden layer, avoiding the uncertainty of random input weights. Based on KPIL, a deep convolutional kernel pseudoinverse learner with a multi-filter design is proposed, named kernel pseudoinverse learning convolutional neural network (KPIL-CNN), which uses multiple fixed convolutional kernels: kernel pseudoinverse learning filters (KPIL filter), Gabor filters, and random filters. The features obtained by this multi-filter approach are combined into feature maps by learning weights. The experimental results show that the proposed KPIL-CNN performs better than other network models with the same scale in the task of natural image classification. The training of KPIL-CNN is efficient and effective. Furthermore, KPIL-CNN outperforms the existing methods in the classification of the ISIC 2017 skin lesion dataset.

Artificial Intelligence and Neural Network-Based Shooting Accuracy Prediction Analysis in Basketball

In order to improve the accuracy of shooting in basketball. A shooting accuracy prediction method based on the convergent improved resource allocating network (CIRAN) online radial basis function neural network (RBFNN) is proposed, and the RBFNN learning algorithm is improved. Through the collection of shooting motion images, feature point extraction, and edge contour feature extraction, the shooting motion trajectory is obtained. Using the online neural network based on the CIRAN learning algorithm to predict the accuracy of shooting, this method analyzes the radial basis function (RBF) network. Based on the RBF analysis, the number of network layers and the number of hidden layer neurons are adjusted and optimized. In order to improve the prediction accuracy of shooting in basketball, a method based on. Through the analysis, it can be known that the accuracy of both the traditional RBFNN and the CIRAN-based online neural network for the prediction of shooting accuracy is above 70%. The prediction accuracy of the online neural network for shooting is higher than that of the traditional one. This is mainly because the online update function of the learning algorithm can better adjust the corresponding structure with the development of the game and has a better generalization ability. In addition, because the CIRAN learning algorithm introduces the hidden layer neuron deletion strategy, its network structure is simpler than that of the traditional one, the number of hidden layer neurons is less, and the running time required is less, which can better meet the real-time requirements and provide a more scientific method for basketball training.

Mine Vegetation Identification via Ecological Monitoring and Deep Belief Network.

Based on the characteristics of remote sensing images of mine vegetation, this research studied the application of deep belief network model in mine vegetation identification. Through vegetation identification and classification, the ecological environment index of mining area was determined according to the analysis of vegetation and coverage. Deep learning algorithm is adopted to improve the depth study, the vegetation coverage in the analysis was studied. Parameters and parameter values were selected for identification by establishing the optimal experimental design. The experimental results were compared with remote sensing images to determine the accuracy of deep learning identification and the effectiveness of the algorithm. When the sample size is 2,000,000 pixels, through repeated tests and classification effect comparison, the optimal parameter setting suitable for mine vegetation identification is obtained. Parameter setting: the number of network layers is 3 layers; the number of hidden layer neurons is 60. The learning rate is 0.01 and the number of iterations is 2. The average recognition rate of vegetation coverage was 95.95%, outperforming some other models, and the accuracy rate of kappa coefficient was 0.95, which can accurately reflect the vegetation coverage. The clearer the satellite image is, the more accurate the recognition result is, and the accuracy is closer to 100%. The identification of vegetation coverage has important guiding significance for determining the area and area of ecological restoration.

Weighted regularized extreme learning machine to model the discharge coefficient of side slots

In the current research, a modern learning machine algorithm named “Weighted Regularized Extreme Learning Machine (WRELM)" is implemented for the first time for the simulation of the coefficient of discharge of side slots. For this purpose, an effective variable on the coefficient of discharge of side slots is firstly introduced, then five distinctive WRELM models are produced by it for the estimation of the coefficient. In the next stage, a database is created for verification of WRELM results. it should be mentioned that 70% of the data are utilized for training the WRELM models, while the rest (i.e. 30%) for testing them. After that, the optimal number of hidden layer neurons as well as the best activation function of the WRELM algorithm are chosen. In addition, the best regularization parameter and also the weight function of the WRELM are achieved. By conducting a sensitivity analysis, the most effective variable for the simulation of the coefficient of discharge along with the WRELM superior model is introduced. The WRELM superior model estimates values of the coefficient of discharge with the maximum exactness and the highest correlation. For instance, the estimations of the correlation coefficient and scatter index for this model are computed to be 0.930 and 0.051, respectively. The sensitivity analysis shows that the ratio of the side slot crest height to its length and the Froude number should be considered as the most important input variables. A comparison between the WRELM with the ELM displays that the former works much better. Furthermore, an uncertainty analysis is executed for both models. Eventually, an equation is suggested for the estimation of the coefficient of discharge and a partial derivative sensitivity analysis is performed on it.

Polynomial Fitting Algorithm Based on Neural Network

As a method of function approximation, polynomial fitting has always been the main research hotspot in mathematical modeling. In many disciplines such as computer, physics, biology, neural networks have been widely used, and most of the applications have been transformed into fitting problems using neural networks. One of the main reasons that neural networks can be widely used is that it has a certain sense of universal approximation. In order to fit the polynomial, this paper constructs a three-layer feedforward neural network, uses Taylor series as the activation function, and determines the number of hidden layer neurons according to the order of the polynomial and the dimensions of the input variables. For explicit polynomial fitting, this paper uses non-linear functions as the objective function, and compares the fitting effects under different orders of polynomials. For the fitting of implicit polynomial curves, the current popular polynomial fitting algorithms are compared and analyzed. Experiments have proved that the algorithm used in this paper is suitable for both explicit polynomial fitting and implicit polynomial fitting. The algorithm is relatively simple, practical, easy to calculate, and can efficiently achieve the fitting goal. At the same time, the computational complexity is relatively low, which has certain application value.

A fetal electrocardiogram signal extraction method based on long short term memory network optimized by genetic algorithm

Fetal electrocardiogram signal extraction is of great significance for perinatal fetal monitoring. In order to improve the prediction accuracy of fetal electrocardiogram signal, this paper proposes a fetal electrocardiogram signal extraction method (GA-LSTM) based on genetic algorithm (GA) optimization with long and short term memory (LSTM) network. Firstly, according to the characteristics of the mixed electrocardiogram signal of the maternal abdominal wall, the global search ability of the GA is used to optimize the number of hidden layer neurons, learning rate and training times of the LSTM network, and the optimal combination of parameters is calculated to make the network topology and the mother body match the characteristics of the mixed signals of the abdominal wall. Then, the LSTM network model is constructed using the optimal network parameters obtained by the GA, and the nonlinear transformation of the maternal chest electrocardiogram signals to the abdominal wall is estimated by the GA-LSTM network. Finally, using the non-linear transformation obtained from the maternal chest electrocardiogram signal and the GA-LSTM network model, the maternal electrocardiogram signal contained in the abdominal wall signal is estimated, and the estimated maternal electrocardiogram signal is subtracted from the mixed abdominal wall signal to obtain a pure fetal electrocardiogram signal. This article uses clinical electrocardiogram signals from two databases for experimental analysis. The final results show that compared with the traditional normalized minimum mean square error (NLMS), genetic algorithm-support vector machine method (GA-SVM) and LSTM network methods, the method proposed in this paper can extract a clearer fetal electrocardiogram signal, and its accuracy, sensitivity, accuracy and overall probability have been better improved. Therefore, the method could extract relatively pure fetal electrocardiogram signals, which has certain application value for perinatal fetal health monitoring.

A Comparison between Seven Heuristic Methods to Estimate the Number of Hidden Layer Neurons in a Multi Layer Perceptron Neural Network

Multilayer Perceptron Neural Network (MLPNNs) constructs of input, at least one hidden and output layer. Number of the neurons in the hidden layer affects the NNs performance. It also consider difficult task to overcome. This research, aims to exanimate the performance of seven heuristic methods that have been used to estimate the neurons numbers in the hidden layer. The effectiveness of these methods was verified using a six of benchmark datasets. The number of hidden layer neurons that selected by each heuristic method for every data set was used to train the MLP. The results demonstrate that the number of hidden neurons selected by each method provides different accuracy and stability compared with other methods. The number of neurons selected by Hush method for ine data set was 26 neurons. It’s achieved the best accuracy with 99.90%and lowest accuracy achieved by Sheela method with 67.51% using 4 neurons. Using 22 neurons with 97.97% accuracy Ke, J method received the best result for Ionosphere data set. While the lowest accuracy was 96.95% with 5 neurons achieved by Kayama method.For Iris data set with 8 neurons achieved 97.19 as best accuracy achieved by Hush method. For the same data set the lowest results were 92.33 % using 3 neurons obtained by using Kayama method. For WBC data set 96.40% the best accuracy achieved using Sheela and Kaastra methods using 4 and 7neurons, while Kanellopoulos method achieved the lowest accuracy 94.18% with 7neurons. For Glass dataset, 87.15% was the best obtained accuracy using 18 neurons Hush method and using Wang method 82.27 % with 6 neurons was the lowest accuracy. Finally for PID 75.31% accuracy achieved by Kayama method with 3 neurons, where Kanellopoulos method obtained 72.17% through using 24 neurons.

A Novel Quantum-Inspired Fuzzy Based Neural Network for Data Classification

The performance of the neural network (NN) depends on the various parameters such as structure, initial weight, number of hidden layer neurons, and learning rate. The improvement in classification performance of NN without changing its structure is a challenging issue. This paper proposes a novel learning model called Quantum-inspired Fuzzy Based Neural Network (Q-FNN) to solve two-class classification problems. In the proposed model, NN architecture is formed constructively by adding neurons in the hidden layer and learning is performed using the concept of Fuzzy c-Means (FCM) clustering, where the fuzziness parameter ( <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">m</i> ) is evolved using the quantum computing concept. The quantum computing concept provides a large search space for a selection of <i xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">m</i> , which helps in finding the optimal weights and also optimizes the network architecture. This paper also proposes a modified step activation function for the formation of hidden layer neurons, which handles the overlapping samples belong to different class regions. The performance of the proposed Q-FNN model is superior and competitive with the state-of-the-art methods in terms of accuracy, sensitivity, and specificity on 15 real-world benchmark datasets.

Optimized extreme learning machine for detecting DDoS attacks in cloud computing

Distributed denial of service (DDoS) attack is a serious security threat to cloud computing that affects the availability of cloud services. Therefore, defending against these attacks becomes imperative. In this paper, we present a DDoS attack detection system based on an improved Self-adaptive evolutionary extreme learning machine (SaE-ELM). SaE-ELM model is improved by incorporating two more features. Firstly, it can adapt the best suitable crossover operator. Secondly, it can automatically determine the appropriate number of hidden layer neurons. These features improve the learning and classification capabilities of the model. The proposed system is evaluated using four datasets namely, NSL-KDD, ISCX IDS 2012, UNSW-NB15, and CICIDS 2017. It achieves the detection accuracy of 86.80%, 98.90%, 89.17%, and 99.99% with NSL-KDD, ISCX IDS 2012, UNSW-NB15, and CICIDS 2017 datasets, respectively. The experiments show that the performance of the proposed attack detection system is better than the system based on original SaE-ELM and state-of-the-art techniques. However, it shows a longer training time than SaE-ELM based system.

Removal of Acid Dye from Aqueous Solutions with Adsorption onto Modified Wheat Bran – Modeling with Artificial Neural Networks

In this study, the removal of 2-Naphthol Orange dye known as Acid Orange 7 (AO7) from aqueous solutions was studied using a modified adsorbent of wheat bran. The efficiency of neural networks in predicting biological uptake was also investigated. The effect of initial concentration, adsorbent dose, pH, velocity, and mixing time in a discontinuous reactor on the adsorption of 2-naphthol orange dye was investigated. Part of the laboratory results was amplified by a feed-forward-back-propagation neural network, and another part was simulated to measure the accuracy of the model. The transmission function and a number of hidden layer neurons were optimized. Optimum conditions were obtained at an initial concentration of 250 mg/L, an adsorbent dose of 0.07 g/L, a pH of 3, a mixing rate of 200 rpm, and a mixing time of 120 min, with a maximum removal rate of 99% and a maximum adsorption capacity of 92.64 mg/g. The dye adsorption kinetics and adsorption equilibrium (isotherm) were found to follow the quasi-second order and Freundlich model, respectively. In the designed neural network, the best transfer function in the sigmoid tangent function's hidden and output layers and the number of optimal neurons was determined to be 13. The model output had a proper correlation with the target vector (R = 0.986). Also, the simulations performed by the neural network model were in good agreement with the experimental results.

Pattern Recognition of Development Stage of Creepage Discharge of Oil–Paper Insulation under AC–DC Combined Voltage Based on OS-ELM

The recognition of the creepage discharge development process of oil–paper insulation under AC–DC combined voltage is the basis for fault monitoring and diagnosis of converter transformers; however, only a few related studies are available. In this study, the AC–DC combined voltage with a ratio of 1:1 was used to develop a recognition method for the creepage discharge development process of an oil–paper insulation under a cylinder–plate electrode structure. First, the pulse current method was used to collect the discharge signals in the creepage discharge development process. Then, 24 characteristic parameters were extracted from four types of creepage discharge characteristic spectra after dimensionality reduction. Finally, based on the online sequential extreme learning machine (OS-ELM) algorithm, these characteristic parameters were used to recognize the development stage of the creepage discharge of the oil–paper insulation. The results showed that when the size of the sample training set used in the OS-ELM algorithm is close to the number of hidden layer neurons, a high recognition accuracy can be obtained, and the type of activation function has little influence on the recognition accuracy. Four stages of the creepage discharge development process were recognized using the OS-ELM algorithm; the trend was the same as that of the characteristic parameters of the entire creepage discharge development process. The recognition accuracy was 91.4%. The algorithm has a high computing speed and high accuracy and can train data in batches. Therefore, it can be widely used in the field of online monitoring and evaluation of electrical equipment status.

Production Forecast Analysis of BP Neural Network Based on Yimin Lignite Supercritical Water Gasification Experiment Results

In the future coal gasification industrial process, it is an efficient method to accurately predict the output of gasification products based on some key industrial parameters. The supercritical water gasification (SCWG) experiment of Yimin lignite was carried out in this work. According to the experimental results, the three main variables of temperature, coal slurry concentration, and residence time are discussed separately for the effects of SCWG products of lignite, and it is proved that temperature is the main influencing parameter of SCWG of lignite. In addition, the experimental results are divided into training set and test set, regression analysis is carried out in BP neural network, and the influence of the number of hidden layer and the number of hidden layer neurons on the result is discussed. The fitting results show that for the prediction of lignite SCWG products, the single-layer neural network has a better fitting effect than the two-layer neural network. The optimized neural network model predicts coal gasification products. The coefficient of determination (R 2 ) of the four products is as high as 0.9908, which is far superior to traditional linear regression.

Prediction of air particulate matter in Beijing, China, based on the improved particle swarm optimization algorithm and long short-term memory neural network

Intelligent algorithms using deep learning can help learn feature data with nonlinearity and uncertainty, such as time-series particle concentration data. This paper proposes an improved particle swarm optimization (IPSO) algorithm using nonlinear decreasing weights to optimize the hyperparameters, such as the number of hidden layer neurons, learning rate, and maximum number of iterations of the long short-term memory (LSTM) neural network, to predict the time series for air particulate concentration and capture its data dependence. The IPSO algorithm uses nonlinear decreasing weights to make the inertia weights nonlinearly decreasing during the iteration process to improve the convergence speed and capability of finding the global optimization of the PSO. This study addresses the limitations of the traditional method and exhibits accurate predictions. The results of the improved algorithm reveal that the root means square, mean absolute percentage error, and mean absolute error of the IPSO-LSTM model predicted changes in six particle concentrations, which decreased by 1.59% to 5.35%, 0.25% to 3.82%, 7.82% to 13.65%, 0.7% to 3.62%, 0.01% to 3.55%, and 1.06% to 17.21%, respectively, compared with the LSTM and PSO-LSTM models. The IPSO-LSTM prediction model has higher accuracy than the other models, and its accurate prediction model is suitable for regional air quality management and effective control of the adverse effects of air pollution.