Neural Network Modeling Method Research Articles

Complex Human Action Recognition Using a Hierarchical Feature Reduction and Deep Learning-Based Method

Automated human action recognition is one of the most attractive and practical research fields in computer vision. In such systems, the human action labelling is based on the appearance and patterns of the motions in the video sequences; however, majority of the existing research and most of the conventional methodologies and classic neural networks either neglect or are not able to use temporal information for action recognition prediction in a video sequence. On the other hand, the computational cost of a proper and accurate human action recognition is high. In this paper, we address the challenges of the preprocessing phase, by an automated selection of representative frames from the input sequences. We extract the key features of the representative frame rather than the entire features. We propose a hierarchical technique using background subtraction and HOG, followed by application of a deep neural network and skeletal modelling method. The combination of a CNN and the LSTM recursive network is considered for feature selection and maintaining the previous information; and finally, a Softmax-KNN classifier is used for labelling the human activities. We name our model as “Hierarchical Feature Reduction & Deep Learning”-based action recognition method, or HFR-DL in short. To evaluate the proposed method, we use the UCF101 dataset for the benchmarking which is widely used among researchers in the action recognition research field. The dataset includes 101 complicated activities in the wild. Experimental results show a significant improvement in terms of accuracy and speed in comparison with eight state-of-the-art methods.

Neural networks and regression analysis in the diagnosis of breast cancer

This work is devoted to the study of the dependence of the temperature fields of the mammary glands on external conditions and the parameters of the anamnesis, and preliminary examination of patients. As a result, it was possible to significantly improve the space of thermometric diagnostic signs intended for the intelligent system. The initial set of highly informative diagnostic thermometric signs was earlier obtained by A. G. Losev and V. V. Levshinsky. To take into account the influence of external factors on the temperature during the formation of the feature space, regression models were proposed. They were built by the method of neural network modeling. These models have sufficient performance and low error value, which allows them to be used in practice. The use of neural networks made it possible to scale the database of thermometric data obtained using a combined and EMC-sensor. As a consequence, it became possible to analyze the influence of the previously revealed heterogeneity of data in the context of age and diameter of the mammary glands on the effectiveness of highly informative diagnostic signs.

Adaptive Online State of Charge Estimation of EVs Lithium-Ion Batteries with Deep Recurrent Neural Networks

The State of Charge (SOC) estimation is a significant issue for safe performance and the lifespan of Lithium-ion (Li-ion) batteries. In this paper, a Robust Adaptive Online Long Short-Term Memory (RoLSTM) method is proposed to extract SOC estimation for Li-ion Batteries in Electric Vehicles (EVs). This real-time, as its name suggests, method is based on a Recurrent Neural Network (RNN) containing Long Short-Term Memory (LSTM) units and using the Robust and Adaptive online gradient learning method (RoAdam) for optimization. In the proposed architecture, one sequential model is defined for each of the three inputs: voltage, current, and temperature of the battery. Therefore, the three networks work in parallel. With this approach, the number of LSTM units are reduced. Using this suggested method, one is not dependent on precise battery models and can avoid complicated mathematical methods. In addition, unlike the traditional recursive neural network where content is re-written at any time, the LSTM network can decide on preserving the current memory through the proposed gateways. In that case, it can easily transfer this information over long paths to receive and maintain long-term dependencies. Using real databases, the experiment results illustrate the better performance of RoLSTM applied to SOC estimation of Li-Ion batteries in comparison with a neural network modeling and unscented Kalman filter method that have been used thus far.

Artificial neural network modeling method incorporating spatial structure

Artificial neural network modeling method incorporating spatial structure

An approach to assessing the technical condition of elements and nodes of transport systems using neural network modeling methods and digital twin technology

The development of intelligent methods and the development of tools for solving research problems of modeling and diagnosing the state of technical systems are relevant areas related to the introduction of digital technologies. Such problems include the problems of preparing correct data arrays for diagnosing and predicting the state of elements and nodes of transport systems. The purpose of the paper is to develop methods for preparing and analyzing data for modeling digital twins on the example of studying the temperature regime of the functioning of a railway car's axle box node, as well as choosing the architecture of neural networks for solving problems of assessing the technical condition of the axle box nodes. The features of the digital twin technology are characterized and the directions of its use for modeling and diagnostics of railway transport systems are considered. As part of the development of an approach to solving the problem of diagnosing the condition of the axle box units of railway cars, analytical dependences and generalized heat transfer equations related to changes in temperature regimes are considered. Examples of the presentation of such data on temperature series, which contain values for deviations from the normal functioning of the elements of the axle box node, are given. The choice of the neural network architecture adapted for solving the problems of estimating and predicting the temperature values of the axle box unit of the car was made. Options for preparing test data for a neural network model have been developed. The results can be used in the problems of creating algorithmic and software for preparing correct arrays of input data for technical diagnostics, in the problems of synthesis and analysis of models of intelligent systems, in various machine learning problems. The considered approach to modeling is aimed at developing methods for assessing and predicting the state of transport infrastructure elements and can be used in the development of intelligent transport systems and for improving the technology of digital twins.

AUTOMATED TELEMEDICAL SYSTEM FOR PRE-HOSPITAL DIAGNOSTICS OF DISEASES AND SUPPORT FOR MAKING DECISIONS BASED ON MACHINE PROCESSING AND ANALYSIS OF IMAGES WITH THE USE OF MOSETHETHINE

The results of the development and features of an automated telemedicine system for pre-hospital diagnostics of diseases and support of medical decision-making based on machine processing and image analysis using neural network modeling methods are considered.

Application of Neural Network Model Based on Multispecies Evolutionary Genetic Algorithm to Planning and Design of Diverse Plant Landscape.

In order to explore the feasibility of applying neural network model to landscape planning, based on the multispecies evolutionary genetic algorithm, a neural network model is proposed in this paper for the system design of diverse plant landscape planning. From the perspective of plant species diversity, this paper discusses landscape planning based on a neural network model. This landscape plan involves more than 180 plant species, mainly shrubs, fungi, and so on. The application of multispecies evolutionary genetic algorithm to landscape planning and design and the application of gene level coding and multispecies parallel evolution strategy to the evolutionary design of neural network have guiding significance for plant landscape planning and design. Compared with the traditional neural network modeling method and genetic algorithm, the proposed method has the advantages of wide network structure search space and simple algorithm calculation and design, independent of specific application background, and has strong application and promotion value. This method makes the model performance evaluation index more comprehensive and accurate and the model solution more reasonable. At the same time, combined with the specific status and corresponding changes of various plants in each season, this paper designs a targeted plan to rationally plan the specific spatial layout of the plant landscape and the combination of different types of plant landscapes, so as to effectively improve the quality of the landscape.

A COMPARISON STUDY BETWEEN MODELING THE HEAT AFFECTED ZONE (HAZ) FOR THE LASER CUTTING OF TI-6AL-4V SHEETS BY USING THE ARTIFICIAL NEURAL NETWORK METHOD AND MULTI REGRESSION METHOD.

This research presents an attempt to study the influence of laser cutting parameters which include Thickness, Lens Focal Length, Beam Power, Cutting Speed and Gas Pressure on the parameter of Heat Affected Zone (HAZ) width. Two predictive models have been developed using Artificial Neural Network (ANN) and multi regression modeling method. The relative importance of laser cutting parameters on (HAZ) width was determined based on (ANN) neuron weights and (ANOVA) method. The comparison between the experimental data with the predicted data indicated that the (ANN) model has attain accuracy of predicting (HAZ) upper than the multi regression model with a coefficient of determination (R2)=85.02%.

Neural network modelling for determining the priority areas of regional development

Neural network modeling based on self-organizing Kohonen maps was carried out in order to cluster the economy of one of the administrative districts of the Krasnoyarsk Territory. The research is based on statistical data on the economy of about one hundred enterprises of various spheres operating in the region. It is shown that in the process of clustering four groups of enterprises can be clearly distinguished according to the types of economic activity. Having applied the neural network modelling method, we identified three enterprises within the framework of this cluster, which can be considered as “growth areas” of the district economy. Self-organizing maps that is trained using unsupervised learning make it possible to get an idea not only of promising areas of development, but also to identify key parameters that ensure the leadership and competitiveness of the region.

Mathematical model for predicting the formation of ferrous deposits during coal combustion

For the first time, on the basis of a combined model of the principal component method and neural network modeling method, a unified method for evaluating coals for the formation of strong ferrous deposits is proposed, taking into account the ash composition, combustion conditions and the degree of grinding. The results of predicting the propensity to form ferrous deposits of Rfe have shown acceptable accuracy of determination. Expanding the experimental database on different types of coals will improve the accuracy of modeling and the widespread use of this technique in the energy sector.

Application of intelligent technologies and neural network modeling methods in the development of a hybrid learning environment

We consider the problems of applying mathematical modeling and intelligent technologies in the development of methods and tools to support educational processes. As a part of creating a hybrid intelligent learning environment, we offer the construction of differential mathematical models with delayed feedback. We characterize a generalized model of the educational process. This model contains a knowledge base module and an interaction module based on artificial intelligence. We study a number of qualitative properties for the dynamic «director–teacher–student» task transfer model with delayed feedback and intelligent components. The described approach to analyzing model properties allows us to consider uncertainties and control actions in hybrid learning environments.

Short-Term Intersection Traffic Flow Forecasting

The intersection is a bottleneck in an urban roadway network. As traffic demand increases, there is a growing congestion problem at urban intersections. Short-term traffic flow forecasting is crucial for advanced trip planning and traffic management. However, there are only a handful of existing models for forecasting intersection traffic flow. In addition, previous short-term traffic flow forecasting models usually were for predicting roadway conditions in a very short period, such as one minute or five minutes, which is often too late given that a driver may well be approaching the bottleneck already. Being able to accurately predict traffic congestions in about half-hour advance is very critical for advanced trip planning and traffic management. To fill this gap, this research develops a two-layer stacking model for intersection short-term traffic flow forecasting by integrating the K-nearest neighbor (KNN) and Elman Neural Network modeling methods. It was developed using the 24-h cycle by cycle traffic data collected at a signalized intersection in Jinan, China. The developed model is evaluated by applying it to the same intersection for forecasting the short-term traffic conditions in a different set of days. The prediction performance of this model was compared with four other models developed using some existing non-parametric modeling and machine learning methods, including clustering, backpropagation (BP) neural network, KNN, and Elman Neural Network. The results of this study indicate that the proposed model outperforms other existing models in terms of its prediction accuracy.

Multi-Fidelity Aerodynamic Data Fusion with a Deep Neural Network Modeling Method

To generate more high-quality aerodynamic data using the information provided by different fidelity data, where low-fidelity aerodynamic data provides the trend information and high-fidelity aerodynamic data provides value information, we applied a deep neural network (DNN) algorithm to fuse the information of multi-fidelity aerodynamic data. We discuss the relationships between the low-fidelity and high-fidelity data, and then we describe the proposed architecture for an aerodynamic data fusion model. The architecture consists of three fully-connected neural networks that are employed to approximate low-fidelity data, and the linear part and nonlinear part of correlation for the low- and high-fidelity data, respectively. To test the proposed multi-fidelity aerodynamic data fusion method, we calculated Euler and Navier–Stokes simulations for a typical airfoil at various Mach numbers and angles of attack to obtain the aerodynamic coefficients as low- and high-fidelity data. A fusion model of the longitudinal coefficients of lift and drag was constructed with the proposed method. For comparisons, variable complexity modeling and cokriging models were also built. The accuracy spread between the predicted value and true value was discussed for both the training and test data of the three different methods. We calculated the root mean square error and average relative deviation to demonstrate the performance of the three different methods. The fusion result of the proposed method was satisfactory on the test case, and showed a better performance compared with the other two traditional methods presented. The results provide evidence that the method proposed in this paper can be useful in dealing with the multi-fidelity aerodynamic data fusion problem.

New optimization design method for a double secondary linear motor based on R-DNN modeling method and MCS optimization algorithm

Traditional linear motor optimization methods typically use analytical models combined with intelligent optimization algorithms. However, this approach has disadvantages, e.g., the analytical model might not be accurate enough, and the intelligent optimization algorithm can easily fall into local optimization. A new linear motor optimization strategy combining an R-deep neural network (R-DNN) and modified cuckoo search (MCS) is proposed; additionally, the thrust lifting and thrust fluctuation reductions are regarded as optimization objectives. The R-DNN is a deep neural network modeling method using the rectified linear unit (RELU) activation function, and the MCS provides a faster convergence speed and stronger data search capability as compared with genetic algorithms, particle swarm optimization, and standard CS algorithms. Finally, the validity and accuracy of this work are proven based on prototype experiments.

A Time Delay Neural Network Based Technique for Nonlinear Microwave Device Modeling.

This paper presents a nonlinear microwave device modeling technique that is based on time delay neural network (TDNN). The proposed technique can accurately model the nonlinear microwave devices when compared to static neural network modeling method. A new formulation is developed to allow for the proposed TDNN model to be trained with DC, small-signal, and large signal data, which can enhance the generalization of the device model. An algorithm is formulated to train the proposed TDNN model efficiently. This proposed technique is verified by GaAs metal-semiconductor-field-effect transistor (MESFET), and GaAs high-electron mobility transistor (HEMT) examples. These two examples demonstrate that the proposed TDNN is an efficient and valid approach for modeling various types of nonlinear microwave devices.

Passenger Flow Prediction Based on Land Use around Metro Stations: A Case Study

High-density land uses cause high-intensity traffic demand. Metro as an urban mass transit mode is considered as a sustainable strategy to balance the urban high-density land uses development and the high-intensity traffic demand. However, the capacity of the metro cannot always meet the traffic demand during rush hours. It calls for traffic agents to reinforce the operation and management standard to improve the service level. Passenger flow prediction is the foremost and pivotal technology in improving the management standard and service level of metro. It is an important technological means in ensuring sustainable and steady development of urban transportation. This paper uses mathematical and neural network modeling methods to predict metro passenger flow based on the land uses around the metro stations, along with considering the spatial correlation of metro stations within the metro line and the temporal correlation of time series in passenger flow prediction. It aims to provide a feasible solution to predict the passenger flow based on land uses around the metro stations and then potentially improving the understanding of the land uses around the metro station impact on the metro passenger flow, and exploring the potential association between the land uses and the metro passenger flow. Based on the data source from metro line 2 in Qingdao, China, the perdition results show the proposed methods have a good accuracy, with Mean Absolute Percentage Errors (MAPEs) of 11.6%, 3.24%, and 3.86 corresponding to the metro line prediction model with Categorical Regression (CATREG), single metro station prediction model with Artificial Neural Network (ANN), and single metro station prediction model with Long Short-Term Memory (LSTM), respectively.

Approaches for Assessing the Polymetallic Impact on Humans in an Urbanized Area Using Neural Network Modeling Methods

A methodology for assessing the response to the polymetallic effects of sensitive population groups has been developed using new information methods and data collection paths used for evaluation. The presented methodology takes into account the processes of polymeric intake, accumulation in the environment, and biosubstrates of the human body. The approach is presented for simultaneous accounting and analysis of dissimilar and heterogeneous data having different degrees of influence on the content of metals in human biosubstrates and the investigated components of the environment using neural self-organizing network Kohonen, implemented in the author’s model of neuronetwork filtration. A method of determining probability of exceeding the threshold values of metal concentrations in the objects under investigation using the risk theory is proposed. As the thresholds of these events, we have taken the median of continuous series of observations of metal concentrations. The facts of established exceeding the thresholds (events) were recorded in the whole set of observations, allocated in the form of clusters formed by the help of neural network classification. In this case, the spatial distribution of selected observations having specific address mappings forms the corresponding territorial zones, which can be used for operational and planned quality management of environmental components in an urbanized territory.

Neural network modeling and combined compromise solution (CoCoSo) method for optimization of drilling performances in polymer nanocomposites

Nowadays, polymer nanocomposite becomes a suitable alternative to conventional materials for lightweight and structural applications. Multiwall carbon nanotube (MWCNT)-reinforced epoxy composites possess superior electrical, mechanical, and thermal properties. In this article, drilling of MWCNT/epoxy nanocomposites has been investigated by considering varying parameters, namely, reinforcement wt% of MWCNT ( W), speed ( N), feed rate ( F), and tool material ( M). The response surface methodology (RSM) array was used for drilling experimentation. The mathematical modeling of drilling parameters was done by using artificial neural network. This study also highlights the integrated approach of principal components analysis (PCA)-embedded combined compromise solution method for multiobjective optimization of conflicting responses such as surface roughness (Ra), torque (Tr), and thrust force (Th). The PCA tool efficiently identified the response priority weight during the aggregation process. The confirmatory test was directed to evaluate the efficiency of the proposed hybrid module. The outcomes show a good agreement between the predicted and experimental value, and it can be endorsed to the polymer manufacturing sector for quality and productivity enhancements.

Neural network modeling methods for predicting the air parameters in the city of Tuzla

According to the report of the World Health Organization, the city of Tuzla is the second in the world, and the first in Europe in terms of the number of diseases caused by air pollution. Tuzla Canton since 2003 has continuous air monitoring. Concentrations of individual pollutants exceed hourly, daily and annual limit values. In this paper, based on the existing results of air monitoring and meteorological data, using statistical methods and neural network modeling methods, unique and reliable models for predicting the concentration of NO2 in the air for the City of Tuzla have been developed. The results obtained using these models can be used in strategic decision-making processes and activities related to air quality control and management. This paper, on the example of the City of Tuzla, showed that using existing air monitoring data, concentrations of pollutants can be predicted for a longer period of time, using artificial intelligence methods. Reliable models with a high correlation coefficient can be obtained. In the case of a short or long interruption of the measurement of pollutant concentrations for the City of Tuzla with the help of models, which are the result of this work, it is possible to predict the concentrations of pollutants and plan to take measures based on them.

О применении нейронных сетей для оценки финансовой устойчивости компаний

The stability of the activities of organizations in the real sector of the economy at the macroeconomic level is the basis for creating the country's gross domestic product, developing technologies and strengthening the competitiveness of the country's products within the global economic system. To plan and forecast their activities, States and business participants are increasingly using modern methods of assessing the financial stability of companies. Today, approaches to building models for assessing the sustainability of companies are based mainly on econometric and statistical linear multidimensional methods of calculation, which does not allow us to identify hidden and nonlinear relationships that are inherent in the real world economy and the activities of economic entities. The article considers approaches to assessing the financial condition of organizations using neural network modeling and comparing it with previously used methods. In the course of the research, we developed a neural network for evaluating the financial condition of companies in the real sector of the economy, which allowed us to draw conclusions about the validity of this method of assessment in modern conditions. The article also highlights the key advantages and disadvantages of the neural network modeling method as a financial analysis tool. The scientific novelty of the article is to develop and evaluate a financial analysis tool that can be applied for practical purposes the economic entities and the substantiation of the complexity of neural network models in predicting bankruptcy; lack of methodological support; the need to develop special software; the duration of the learning process to achieve the required accuracy of the model; compliance with the requirement for equal proportions of the studied groups of objects; the correct choice of neural network architecture for research purpose; representativeness and consistency of source data.