Search results

The purpose of this paper is to present research in the area of the modeling of complex systems using feed‐forward neural network.

Applications of multilayer neural networks with supervisor learning on the own simulator program wrote in Borland^® Pascal Language. Series‐parallel identification method is applied. Tapped delay lines (TDL) in static neural networks for modeling of dynamic plants are used. Gradient and heuristic learning algorithms are applied. Three kinds of calibration of learning and testing data are used.

This paper illustrates that feed‐forward multilayer neural networks can model complex systems. Feed‐forward multilayer neural networks with TDL can be used to build global dynamic models of complex systems. It is possible to compare the quality both models.

The learning and testing data from real systems to tune neuronal models require use of calibrating these data to range 0‐1.

The models quality depends on kind of calibration learning data from real system and depends on kind of learning algorithms.

The method and the learning algorithms discussed in the paper can be used to create global models of complex systems. The multilayer neural network with TDL can be used to model complex dynamic systems with low dynamics.

The purpose of this paper is to build a neural model of an aircraft from flight data and online estimation of the aerodynamic derivatives from established neural model.

A neural model capable of predicting generalized force and moment coefficients of an aircraft using measured motion and control variable is used to extract aerodynamic derivatives. The use of neural partial differentiation (NPD) method to the multi-input-multi-output (MIMO) aircraft system for the online estimation of aerodynamic parameters from flight data is extended.

The estimation of aerodynamic derivatives of rigid and flexible aircrafts is treated separately. In the case of rigid aircraft, longitudinal and lateral-directional derivatives are estimated from flight data. Whereas simulated data are used for a flexible aircraft in the absence of its flight data. The unknown frequencies of structural modes of flexible aircraft are also identified as part of estimation problem in addition to the stability and control derivatives. The estimated results are compared with the parameter estimates obtained from output error method. The validity of estimates has been checked by the model validation method, wherein the estimated model response is matched with the flight data that are not used for estimating the derivatives.

Compared to the Delta and Zero methods of neural networks for parameter estimation, the NPD method has an additional advantage of providing the direct theoretical insight into the statistical information (standard deviation and relative standard deviation) of estimates from noisy data. The NPD method does not require the initial value of estimates, but it requires a priori information about the model structure of aircraft dynamics to extract the flight stability and control parameters. In the case of aircraft with a high degree of flexibility, aircraft dynamics may contain many parameters that are required to be estimated. Thus, NPD seems to be a more appropriate method for the flexible aircraft parameter estimation, as it has potential to estimate most of the parameters without having the issue of convergence.

This paper highlights the application of NPD for MIMO aircraft system; previously it was used only for multi-input and single-output system for extraction of parameters. The neural modeling and application of NPD approach to the MIMO aircraft system facilitate to the design of neural network-based adaptive flight control system. Some interesting results of parameter estimation of flexible aircraft are also presented from established neural model using simulated data as a novelty. This gives more value addition to analyzing the flight data of flexible aircraft as it is a challenging problem in parameter estimation of flexible aircraft.

Credit scoring is important for financial institutions in order to accurately predict the likelihood of business failure. Related studies have shown that machine learning techniques, such as neural networks, outperform many statistical approaches to solving this type of problem, and advanced machine learning techniques, such as classifier ensembles and hybrid classifiers, provide better prediction performance than single machine learning based classification techniques. However, it is not known which type of advanced classification technique performs better in terms of financial distress prediction. The paper aims to discuss these issues.

This paper compares neural network ensembles and hybrid neural networks over three benchmarking credit scoring related data sets, which are Australian, German, and Japanese data sets.

The experimental results show that hybrid neural networks and neural network ensembles outperform the single neural network. Although hybrid neural networks perform slightly better than neural network ensembles in terms of predication accuracy and errors with two of the data sets, there is no significant difference between the two types of prediction models.

The originality of this paper is in comparing two types of advanced classification techniques, i.e. hybrid and ensemble learning techniques, in terms of financial distress prediction.

Online news websites provide huge amounts of timely news, bringing the challenge of recommending personalized news articles. Generative adversarial networks (GAN) based on collaborative filtering (CFGAN) can achieve effective recommendation quality. However, CFGAN ignores item contents, which contain more latent preference features than just user ratings. It is important to consider both ratings and item contents in making preference predictions. This study aims to improve news recommendation by proposing a GAN-based news recommendation model considering both ratings (implicit feedback) and the latent features of news content.

The collaborative topic modeling (CTM) can improve user preference prediction by combining matrix factorization (MF) with latent topics of item content derived from latent topic modeling. This study proposes a novel hybrid news recommendation model, Hybrid-CFGAN, which modifies the architecture of the CFGAN model with enhanced preference learning from the CTM. The proposed Hybrid-CFGAN model contains parallel neural networks – original rating-based preference learning and CTM-based preference learning, which consider both ratings and news content with user preferences derived from the CTM model. A tunable parameter is used to adjust the weights of the two preference learnings, while concatenating the preference outputs of the two parallel neural networks.

This study uses the dataset collected from an online news website, NiusNews, to conduct an experimental evaluation. The results show that the proposed Hybrid-CFGAN model can achieve better performance than the state-of-the-art GAN-based recommendation methods. The proposed novel Hybrid-CFGAN model can enhance existing GAN-based recommendation and increase the performance of preference predictions on textual content such as news articles.

As the existing CFGAN model does not consider content information and solely relies on history logs, it may not be effective in recommending news articles. Our proposed Hybrid-CFGAN model modified the architecture of the CFGAN generator by adding a parallel neural network to gain the relevant information from news content and user preferences derived from the CTM model. The novel idea of adjusting the preference learning from two parallel neural networks – original rating-based preference learning and CTM-based preference learning – contributes to improve the recommendation quality of the proposed model by considering both ratings and latent preferences derived from item contents. The proposed novel recommendation model can improve news recommendation, thereby increasing the commercial value of news media platforms.

This paper aims to evaluate the performance of the Mashhad No. 5 wastewater treatment plant (WWTP) using a combination of data mining (regression) algorithms and artificial neural networks.

In this research, the performance of WWTP located in Mashhad, Iran, has been evaluated using two data mining models, neural network and regression model.

The proposed model has the potential of implementing in other WWTPs in Iran or other countries.

The authors would also like to thank Mashhad No.5 WWTP for data access.

Estimation of the rental price of a residential property is important to real estate investors, financial institutions, buyers and the government. These estimates provide information for assessing the economic viability and the tax accruable, respectively. The purpose of this study is to develop a neural network model for estimating the rental prices of residential properties in Cape Town, South Africa.

Data were collected on 14 property attributes and the rental prices were collected from relevant sources. The neural network algorithm was used for model estimation and validation. The data relating to 286 residential properties were collected in 2018.

The results show that the predictive accuracy of the developed neural network model is 78.95 per cent. Based on the sensitivity analysis of the model, it was revealed that balcony and floor area have the most significant impact on the rental price of residential properties. However, parking type and swimming pool had the least impact on rental price. Also, the availability of garden and proximity of police station had a low impact on rental price when compared to balcony.

In the light of these results, the developed neural network model could be used to estimate rental price for taxation. Also, the significant variables identified need to be included in the designs of new residential homes and this would ensure optimal returns to the investors.

A number of studies have shown that crime influences the value of residential properties. However, to the best of the authors’ knowledge, there is limited research investigating this relationship within the South African context.

Artificial neural networks (ANNs), which represent computational models simulating the biological neural systems, have become a dominant paradigm for solving complex analytical problems. ANN applications have been employed in various disciplines such as psychology, computer science, mathematics, engineering, medicine, manufacturing, and business studies. Academic research on ANNs is witnessing considerable publication activity, and there exists a need to track the intellectual structure of the existing research for a better comprehension of the domain. The current study uses a bibliometric approach to ANN business literature extracted from the Web of Science database. The study also performs a chronological review using science mapping and examines the evolution trajectory to determine research areas relevant to future research. The authors suggest that researchers focus on ANN deep learning models as the bibliometric results predict an expeditious growth of the research topic in the upcoming years. The findings reveal that business research on ANNs is flourishing and suggest further work on domains, such as back-propagation neural networks, support vector machines, and predictive modeling. By providing a systematic and dynamic understanding of ANN business research, the current study enhances the readers' understanding of existing reviews and complements the domain knowledge.

The purpose of this paper is to propose a new aerodynamic parameter estimation methodology based on neural network and output error method, while the output error method is improved based on particle swarm algorithm.

Firstly, the algorithm approximates the dynamic characteristics of aircraft based on feedforward neural network. Neural network is trained by extreme learning machine, and the trained network can predict the aircraft response at (k + 1)th instant given the measured flight data at kth instant. Secondly, particle swarm optimization is used to enhance the convergence of Levenberg–Marquardt (LM) algorithm, and the improved LM method is used to substitute for the Gauss Newton algorithm in output error method. Finally, the trained neural network is combined with the improved output error method to estimate aerodynamic derivatives.

Neither depending on the initial guess of the parameters to be estimated nor requiring numerical integration of the aircraft motion equation, the proposed algorithm can be used for unstable aircraft and is successfully applied to extract aerodynamic derivatives from both simulated and real flight data.

The proposed method requires iterative calculation and can only identify parameters offline.

The proposed method is successfully applied to estimate aircraft aerodynamic parameters and can also be used as a new algorithm for other optimization problems.

In this study, the output error method is improved to reduce the dependence on the initial value of parameters and expand its application scope. It is applied in aircraft aerodynamic parameter identification together with neural network.

The purpose of this paper is to improve back propagation neural network (BPNN) modeling in order to promote the forecast calculation precision of landslide deformation.

The genetic algorithm is adopted to optimize the architectural parameter of BPNN so as to avoided errors occurrence while using the trial‐and‐error method. Furthermore, the Sigmoid function is improved and revised to expand the output range of change‐over function from unipolar (only positive) to ambipolar (may be positive or negative), then the convergence time is reduced and the neural network can express more artificial intelligence.

The modeling can effectively reduce the probability to get into the local minima while employing neural networks to forecast the landslide deformation. It significantly promotes the forecast precision.

The improved BPNN modeling, which is very good in learning and processing information, can work out the complex non‐linear relation by learning model and using the present data or reciprocity of surroundings.

The revised BPNN modeling in this paper can be used to predict and calculate landslide deformation.

The paper demonstrates that the modeling can meet the demand of calculation precision.

This paper seeks to present a novel computational intelligence technique to generate concise neural network models for distributed dynamic systems.

The approach used in this paper is based on artificial neural network architectures that incorporate linear and nonlinear principal component analysis, combined with generalized dimensional analysis.

Neural network principal component analysis coupled with generalized dimensional analysis reduces input variable space by about 90 percent in the modeling of oil reservoirs. Once trained, the computation time is negligible and orders of magnitude faster than any traditional discretisation schemes such as fine‐mesh finite difference.

Finding the minimum number of input independent variables needed to characterize a system helps in extracting general rules about its behavior, and allows for quick setting of design guidelines, and particularly when evaluating changes in the physical properties of systems.

The methodology can be used to simulate dynamical systems characterized by differential equations, in an interactive CAD and optimization providing faster on‐line solutions and speeding up design guidelines.