Individual Artificial Neural Network Research Articles

Assessment of soil thermal conduction using artificial neural network models

Thermal conductivity is a fundamental engineering property governing heat transfer process in soils. It depends on mineral component, compaction moisture content, dry density, soil gradation and temperature, and usually varies over an order of magnitude in soils with different status. The present study investigated the mechanism of heat transfer in soils and developed new models for thermal conductivity prediction via artificial neural network (ANN) technology. The performance of the proposed models (individual ANN model and generalized ANN model) were evaluated by comparing with three empirical models. Based on the results presented in this study, it is revealed that heat flow through soil was a multi-field coupled process (i.e., thermal-hydro-mechanical process) and was closely related to the intrinsic properties of three phases those constitute a soil. The cross validation was conducted to validate the reliability of the proposed models. It was concluded that both individual and generalized ANN models were able to provide good matching with laboratory measured thermal conductivity values. For each proposed model, the coefficient of correlation (R2) and variance account for (VAF) values were close to 1, and the mean absolute error (MAE) and root mean square error (RMSE) were lower than 0.360 W/K m and 1.000 W/K m, respectively. Results taken from the comparison of various models showed that the generalized ANN model, with RMSE value lower than 0.100 W/K m, exhibited highest accuracy in thermal conductivity prediction of all types of soil, followed by individual ANN models and the empirical models. A good performance of the proposed models in frozen soils was observed with a limited size of thermal conductivity data.

Prediction of electronic structure in atomistic model using artificial neural network

A scheme to predict electronic density of states (DoS) from atomic configurations using the artificial neural network (ANN) is proposed. Integrated DoS (IDoS) was reproduced by discretizing IDoS into and constructing individual ANN elements corresponding to sampling points of the discretization. The validity of this scheme was examined and found successful for typical crystal structures of SiC, C and Si. The proposed scheme paves the way to new approaches to obtain various materials properties without employing electronic state calculations such as first-principles methods.

A fusion-based neural network methodology for monthly reservoir inflow prediction using MODIS products

ABSTRACTA model fusion approach was developed based on five artificial neural networks (ANNs) and MODIS products. Static and dynamic ANNs – the multi-layer perceptron (MLP) with one and two hidden layers, general regression neural network (GRNN), radial basis function (RBF) and nonlinear autoregressive network with exogenous inputs (NARX) – were used to predict the monthly reservoir inflow in Mollasadra Dam, Fars Province, Iran. Leaf area index and snow cover from MODIS, and rainfall and runoff data were used to identify eight different combinations to train the models. Statistical error indices and the Borda count method were used to verify and rank the identified combinations. The best results for individual ANNs were combined with MODIS products in a fusion model. The results show that using MODIS products increased the accuracy of predictions, with the MLP with two hidden layers giving the best performance. Also, the fusion model was found to be superior to the best individual ANNs.

Basin-Scale Prediction of Sea Surface Temperature with Artificial Neural Networks

AbstractThe prediction of sea surface temperature (SST) on the basis of artificial neural networks (ANNs) can be viewed as complementary to numerical SST predictions, and it has fairly sustained in the recent past. However, one of its limitations is that such ANNs are site specific and do not provide simultaneous spatial information similar to the numerical schemes. In this work we have addressed this issue by presenting basin-scale SST predictions based on the operation of a very large number of individual ANNs simultaneously. The study area belongs to the basin of the tropical Indian Ocean (TIO) having coordinates of 30°N–30°S, 30°–120°E. The network training and testing are done on the basis of HadISST data of the past 140 yr. Monthly SST anomalies are predicted at 3813 nodes in the basin and over nine time steps into the future with more than 20 million ANN models. The network testing indicated that the prediction skill of ANNs is attractive up to certain lead times depending on the subbasin. The ANN models performed well over both the western Indian Ocean (WIO) and eastern Indian Ocean (EIO) regions up to 5 and 4 months lead time, respectively, as judged by the error statistics of the correlation coefficient and the normalized root-mean-square error. The prediction skill of the ANN models for the TIO region is found to be better than the physics-based coupled atmosphere–ocean models. It is also observed that the ANNs are capable of providing an advanced warning of the Indian Ocean dipole as well as abnormal basin warming.

Data Pre-Analysis and Ensemble of Various Artificial Neural Networks for Monthly Streamflow Forecasting

This paper introduces three artificial neural network (ANN) architectures for monthly streamflow forecasting: a radial basis function network, an extreme learning machine, and the Elman network. Three ensemble techniques, a simple average ensemble, a weighted average ensemble, and an ANN-based ensemble, were used to combine the outputs of the individual ANN models. The objective was to highlight the performance of the general regression neural network-based ensemble technique (GNE) through an improvement of monthly streamflow forecasting accuracy. Before the construction of an ANN model, data preanalysis techniques, such as empirical wavelet transform (EWT), were exploited to eliminate the oscillations of the streamflow series. Additionally, a theory of chaos phase space reconstruction was used to select the most relevant and important input variables for forecasting. The proposed GNE ensemble model has been applied for the mean monthly streamflow observation data from the Wudongde hydrological station in the Jinsha River Basin, China. Comparisons and analysis of this study have demonstrated that the denoised streamflow time series was less disordered and unsystematic than was suggested by the original time series according to chaos theory. Thus, EWT can be adopted as an effective data preanalysis technique for the prediction of monthly streamflow. Concurrently, the GNE performed better when compared with other ensemble techniques.

A Hybridized Fuzzy-Neural Predictive Intelligent (HFNPI) Modelling Approach-based Underlap FinFET Model

ABSTRACTThis paper presents the hybridized fuzzy-neural predictive intelligent (HFNPI) based approach for predicting FinFET model. The proposed model is an improved approach over artificial neural network (ANN) model and adaptive neuro fuzzy inference system (ANFIS) model that provides optimum model parameters. In this paper, the challenge of optimization of large number of parameters is reduced. It assesses the performance of resultant structure on the basis of selected RF figure-of-merit. Based on the application requirements, RF figure-of-merit such as fT and fmax are used as appropriate performance parameters to analyse the device performance. The framework for the HFNPI model is underlapped FinFET structural/process parameters. The mean prediction error was found 2.8 × 10−4 GHz. The comparative analysis of HFNPI with ANN and ANFIS model is done. It is found that combination of fuzzy and neural logic shows better result rather than individual ANFIS and ANN model to redevelop the model and increase the number of inputs.

Oilseeds production and yield forecasting using ARIMA-ANN modelling

Agriculture, the backbone of Indian economy, lot of time series data on various parameters and these time series data generated from agriculture can be effectively modelled using various time-series modelling techniques such as Box-Jenkins ARIMA modelling technique, State-Space modelling technique, Structural Time Series modelling methodology and various other time series modelling methodologies depending upon the availability and properties of the given datasets. Modelling and forecasting of time-series data on all-India production and productivity of nine oilseed crops from 1950-51 to 2015-16 is carried out in this study. Modelling and subsequent forecasting for the datasets under consideration is performed using Autoregressive Moving Average (ARIMA), Artificial Neural Network (ANN) and ARIMA-ANN hybrid modelling methodologies. It is important to note that ARIMA is a linear modelling methodology whereas ANN is a non-linear modelling methodology. ARIMA-ANN is a hybrid of these two methodologies which can efficiently capture both linear and non-linear structures present in the dataset under consideration. For efficient performance of ARIMA modelling, the data under consideration needs to satisfy stationarity criterion. i.e. mean and variance should constant over a period of time. First, the dataset on oilseeds production and productivity are tested for stationarity and subsequent to non-stationarity of the original data, firstorder differenced series are considered for modeling using the Box-Jenkins approach. Various ARIMA models are developed and among them ARIMA (1,1,0) and ARIMA (1,1,1) are found suitable for the production andproductivity data, respectively, based on the Information Criteria such as Akaike Information Criterion (AIC) and Schwarz-Bayesian Criterion (SBC). Among the developed ANN models, the Neural Network Autoregression(NNAR) of order NNAR (2,2) is found to be suitable for both the variables under study. Later, models using ANN-ARIMA hybrid methodology are developed and the ARIMA (1,1,0)-NNAR(1,1) and ARIMA (1,1,1)-NNAR(1,1) for production and productivity are found to be suitable, respectively. All the three models are tested for their forecast accuracy using Mean Absolute Percentage Error (MAPE) and Root Mean Square Error (RMSE). Accordingly, the ARIMA-ANN hybrid methodology is found to be superior to the individual ARIMA and ANN methodologies. Based on the efficient developed model viz., hybrid ARIMA-ANN model, forecasting annual all-India oilseeds production for the year 2022 is carried out and is found to be 35.6 (±3.2) million tonnes with a productivity of 1178 (±94.6) kg/hectare.

A Smart Forecasting Approach to District Energy Management

This study presents a model for district-level electricity demand forecasting using a set of Artificial Neural Networks (ANNs) (parallel ANNs) based on current energy loads and social parameters such as occupancy. A comprehensive sensitivity analysis is conducted to select the inputs of the ANN by considering external weather conditions, occupancy type, main income providers’ employment status and related variables for the fuel poverty index. Moreover, a detailed parameter tuning is conducted using various configurations for each individual ANN. The study also demonstrates the strength of the parallel ANN models in different seasons of the years. In the proposed district level energy forecasting model, the training and testing stages of parallel ANNs utilise dataset of a group of six buildings. The aim of each individual ANN is to predict electricity consumption and the aggregated demand in sub-hourly time-steps. The inputs of each ANN are determined using Principal Component Analysis (PCA) and Multiple Regression Analysis (MRA) methods. The accuracy and consistency of ANN predictions are evaluated using Pearson coefficient and average percentage error, and against four seasons: winter, spring, summer, and autumn. The lowest prediction error for the aggregated demand is about 4.51% for winter season and the largest prediction error is found as 8.82% for spring season. The results demonstrate that peak demand can be predicted successfully, and utilised to forecast and provide demand-side flexibility to the aggregators for effective management of district energy systems.

A double-elimination-tournament-based competitive co-evolutionary artificial neural network classifier

This paper presents a competitive co-evolutionary (ComCoE) that engages a double elimination tournament (DET) to evolve artificial neural networks (ANNs) for undertaking data classification problems. The proposed model performs a global search by a ComCoE approach to find near optimal solutions. During the global search process, two populations of different ANNs compete and fitness evaluation of each ANN is made in a subjective manner based on their participations throughout a DET which promotes competitive interactions among individual ANNs. The adaptation and fitness evaluation processes drive the global search for a more competent ANN classifier. A winning ANN is identified from the global search. Then, the Scaled Conjugate Backpropagation algorithm, which is a local search, is performed to further train the winning ANN to obtain a precise solution. The performance of the proposed classification model is evaluated rigorously; its performance is compared with the baseline ANNs of the proposed model as well as other classifiers. The results indicate that the proposed model could construct an ANN which could produce high classification accuracy rates with a compact network structure.

Systematic analysis of non-structural protein features for the prediction of PTM function potential by artificial neural networks.

Post-translational modifications (PTMs) provide an extensible framework for regulation of protein behavior beyond the diversity represented within the genome alone. While the rate of identification of PTMs has rapidly increased in recent years, our knowledge of PTM functionality encompasses less than 5% of this data. We previously developed SAPH-ire (Structural Analysis of PTM Hotspots) for the prioritization of eukaryotic PTMs based on function potential of discrete modified alignment positions (MAPs) in a set of 8 protein families. A proteome-wide expansion of the dataset to all families of PTM-bearing, eukaryotic proteins with a representational crystal structure and the application of artificial neural network (ANN) models demonstrated the broader applicability of this approach. Although structural features of proteins have been repeatedly demonstrated to be predictive of PTM functionality, the availability of adequately resolved 3D structures in the Protein Data Bank (PDB) limits the scope of these methods. In order to bridge this gap and capture the larger set of PTM-bearing proteins without an available, homologous structure, we explored all available MAP features as ANN inputs to identify predictive models that do not rely on 3D protein structural data. This systematic, algorithmic approach explores 8 available input features in exhaustive combinations (247 models; size 2–8). To control for potential bias in random sampling for holdback in training sets, we iterated each model across 100 randomized, sample training and testing sets—yielding 24,700 individual ANNs. The size of the analyzed dataset and iterative generation of ANNs represents the largest and most thorough investigation of predictive models for PTM functionality to date. Comparison of input layer combinations allows us to quantify ANN performance with a high degree of confidence and subsequently select a top-ranked, robust fit model which highlights 3,687 MAPs, including 10,933 PTMs with a high probability of biological impact but without a currently known functional role.

Multicolumn RBF Network.

This paper proposes the multicolumn RBF network (MCRN) as a method to improve the accuracy and speed of a traditional radial basis function network (RBFN). The RBFN, as a fully connected artificial neural network (ANN), suffers from costly kernel inner-product calculations due to the use of many instances as the centers of hidden units. This issue is not critical for small datasets, as adding more hidden units will not burden the computation time. However, for larger datasets, the RBFN requires many hidden units with several kernel computations to generalize the problem. The MCRN mechanism is constructed based on dividing a dataset into smaller subsets using the k-d tree algorithm. resultant subsets are considered as separate training datasets to train individual RBFNs. Those small RBFNs are stacked in parallel and bulged into the MCRN structure during testing. The MCRN is considered as a well-developed and easy-to-use parallel structure, because each individual ANN has been trained on its own subsets and is completely separate from the other ANNs. This parallelized structure reduces the testing time compared with that of a single but larger RBFN, which cannot be easily parallelized due to its fully connected structure. Small informative subsets provide the MCRN with a regional experience to specify the problem instead of generalizing it. The MCRN has been tested on many benchmark datasets and has shown better accuracy and great improvements in training and testing times compared with a single RBFN. The MCRN also shows good results compared with those of some machine learning techniques, such as the support vector machine and k-nearest neighbors.

Airborne castanea pollen forecasting model for ecological and allergological implementation

Castanea sativa Miller belongs to the natural vegetation of many European deciduous forests prompting impacts in the forestry, ecology, allergological and chestnut food industry fields. The study of the Castanea flowering represents an important tool for evaluating the ecological conservation of North-Western Spain woodland and the possible changes in the chestnut distribution due to recent climatic change. The Castanea pollen production and dispersal capacity may cause hypersensitivity reactions in the sensitive human population due to the relationship between patients with chestnut pollen allergy and a potential cross reactivity risk with other pollens or plant foods. In addition to Castanea pollen's importance as a pollinosis agent, its study is also essential in North-Western Spain due to the economic impact of the industry around the chestnut tree cultivation and its beekeeping interest. The aim of this research is to develop an Artificial Neural Networks for predict the Castanea pollen concentration in the atmosphere of the North-West Spain area by means a 20years data set.It was detected an increasing trend of the total annual Castanea pollen concentrations in the atmosphere during the study period. The Artificial Neural Networks (ANNs) implemented in this study show a great ability to predict Castanea pollen concentration one, two and three days ahead. The model to predict the Castanea pollen concentration one day ahead shows a high linear correlation coefficient of 0.784 (individual ANN) and 0.738 (multiple ANN). The results obtained improved those obtained by the classical methodology used to predict the airborne pollen concentrations such as time series analysis or other models based on the correlation of pollen levels with meteorological variables.

Artificial neural network models for depicting mass flow rate of R22, R407C and R410A through electronic expansion valves

The utilization of electronic expansion valves (EEVs) in refrigeration and air conditioning systems is increased for energy saving and comfort environmental. However, experimental data and refrigerant mass flow models through EEVs are very limited in open literature. In this study, a new technique using artificial neural network (ANN) is applied to depict the mass flow rates of R22 and its alternatives R407C and R410A flowing through EEVs based on the error back propagation learning algorithm. Two strategies are followed; the first is to construct individual ANN models for each refrigerant, and the second is to construct a generalized ANN model for the three investigated refrigerants. The experimental results from open literature are used to construct the ANN models. The ANN models results showed a good agreement with the corresponding experimental data. For individual models, the relative deviations for R22, R407C, and R410A are within ±0.7%, ±1.1%, and ±0.006%, respectively. While for generalized model, the relative deviations are within ±2.5%. Also the generalized model was tested out of its construction range in a predictive mode and it was found to be a reliable tool to estimate the refrigerants mass flow rates.

Hybrid computing techniques for fault detection and isolation, a review

The classical model-based methods had often proven to be unable to provide acceptable solutions to modern fault diagnosis systems. Therefore, model-free or soft computing techniques such as fuzzy logic, Artificial Neural Networks (ANNs) and Genetic Algorithms (GAs) had become more attractive in industrial applications of fault diagnosis. In this paper, three SC schemes are explored to solve the problem of detecting unprecedented changes and finding the failed state components. First, individual fuzzy systems, ANN and GA are implemented on a fault diagnosis scheme. Then, hybrids of these techniques are applied to enhance the fault diagnosis precision. This approach allows gaining critical information about fault presence or its absence in the shortest possible time. The proposed scheme was simulated and evaluated extensively on a benchmark laboratory scale coupled-two-tank system. The results are encouraging, showing especially, that hybrid GA+ANFIS (GANFIS), outperformed significantly the other techniques.

Multiple neural networks switched prediction for landslide displacement

An accurate prediction of landslide displacement is challenging and of great interest to governments and researchers. In order to reduce the risk of selecting the types of influencing factors and artificial neural networks (ANNs), a multiple ANNs switched prediction method is proposed for landslide displacement forecasting. In the first stage, a set of individual neural networks are developed based on different environmental factors and/or different training algorithms. In the second stage, a switched prediction method is used to select the appropriate individual neural network for prediction purpose. For verification and testing, three typical landslides in Three Gorges Reservoir, namely Baishuihe landslide, Bazimen landslide and Shiliushubao landslide, are presented to test the effectiveness of our method. Application results demonstrate that the proposed method can significantly improve model generalization and perform similarly to, or better than, the best individual ANN predictor.

A moving-average filter based hybrid ARIMA–ANN model for forecasting time series data

A suitable combination of linear and nonlinear models provides a more accurate prediction model than an individual linear or nonlinear model for forecasting time series data originating from various applications. The linear autoregressive integrated moving average (ARIMA) and nonlinear artificial neural network (ANN) models are explored in this paper to devise a new hybrid ARIMA–ANN model for the prediction of time series data. Many of the hybrid ARIMA–ANN models which exist in the literature apply an ARIMA model to given time series data, consider the error between the original and the ARIMA-predicted data as a nonlinear component, and model it using an ANN in different ways. Though these models give predictions with higher accuracy than the individual models, there is scope for further improvement in the accuracy if the nature of the given time series is taken into account before applying the models. In the work described in this paper, the nature of volatility was explored using a moving-average filter, and then an ARIMA and an ANN model were suitably applied. Using a simulated data set and experimental data sets such as sunspot data, electricity price data, and stock market data, the proposed hybrid ARIMA–ANN model was applied along with individual ARIMA and ANN models and some existing hybrid ARIMA–ANN models. The results obtained from all of these data sets show that for both one-step-ahead and multistep-ahead forecasts, the proposed hybrid model has higher prediction accuracy.

WE‐C‐108‐03: CT‐Based 3D Dose Calculation Method Using Artificial Neural Networks (ANN)

Purpose: In this study Artificial Neural Networks (ANN) are trained to develop a fast CT based dose calculation method. The ANN is capable of calculating the 3D dose distribution inside a voxel‐based phantom for prostate I–125 seed implants. The speed allows for accurate CT‐based calculation of the dose distributions during implantation in real time due to plan modifications. Methods: Training the ANN needs a set of accurate sample dose distributions in heterogeneous CT based phantoms. For the purpose of this study, 60000 sample cubicle phantoms were extracted from 3 patient CT studies. The dose distributions of the cubicle phantoms were calculated using MCNP5. The absorption coefficient of each voxel of the cubicle phantom was calculated by analyzing the CT scan. This information was then given to the ANN as the input. The calculated dose by MCNP5 was given to it as the sample output for the purpose of training. The dose distributions calculated by ANNs were then tested for 7000 cubicle phantoms against the MCNP5 results. Finally, the ANN was used to calculate the dose distribution in two real treatment plans. The results were compared against the MCNP5 voxel based results. Results: The regression graphs for individual ANNs show a regression of R=0.99543 for the worst ANN The regression of the best Trained ANN was equal to R=0.9997 for the 7000 cases. The maximum average absolute error of the ANN dose calculation was 3% in comparison with MCNP5. Conclusion: The results of individual ANN calculations and the complete treatment plan calculations verified the accuracy of the ANN method. Capability of the ANN method to calculate the dose distributions for 500 seed positions in less than 10 minutes suggests that this method can be used for model based dose calculation for real‐time plan modifications in the case of LDR brachytherapy.

Identification of member connectivity and mass changes on a two-storey framed structure using frequency response functions and artificial neural networks

This paper presents a structural health monitoring (SHM) technique that utilises pattern changes in frequency response functions (FRFs) as input parameters for a system of artificial neural networks (ANNs) to assess the structural condition of a structure. To verify the proposed method, it is applied to numerical and experimental models of a two-storey framed structure, on which structural damage is induced by member connectivity and mass changes, respectively. For the numerical structure, simulated time-history data are polluted with various levels of white Gaussian noise in order to realistically represent field-testing conditions. As a damage indicator, residual FRFs are used, which are derived by calculating the differences in FRF data between the undamaged/baseline structure and the structure with changed joint conditions or added mass. To obtain suitable patterns for neural network training, principal component analysis (PCA) techniques are adopted to reduce the size of the residual FRF data and to filter noise. A hierarchical system of individual ANNs, termed network ensemble, is then trained to map changes in PCA-reduced residual FRFs to damage conditions. The results obtained for both damage investigations, namely joint damage and mass changes, demonstrate that the proposed SHM technique is accurate and reliable in assessing the condition of the test structure numerically and experimentally based on direct FRF measurements and network ensemble analysis. From the outcomes of the individual networks, it is found that the proposed hierarchical network ensemble approach is highly efficient in filtering poor results of underperforming networks obtained from measurement locations with low damage sensitivity.

Comparison of individual and combined ANN models for prediction of air and dew point temperature

Predicted air and dew point temperatures can be valuable in decision making in many areas including protecting crops from damage, avoiding heat stress on animals and humans, and in planning related to energy management. Current web-based artificial neural network (ANN) models on the Automated Environment Monitoring Network (AEMN) in Georgia predict hourly air and dew point temperature for twelve prediction horizons, using 24 models. The observed air temperature may approach the observed dew point temperature, but never goes below it. Current web based ANN models have prediction errors which, when the air and dew point temperatures are close, may cause air temperature to be predicted below the dew point temperature. Herein this error is referred to as a prediction anomaly. The goal of this research was to improve the prediction accuracy of existing air and dew point temperature ANN models by combining the two weather variables into a single ANN model for each prediction horizon. The objectives of this study were to reduce the mean absolute error (MAE) of prediction and to reduce the number of prediction anomalies. The combined models produced a reduction in the air temperature MAE for ten of twelve prediction horizons with an average reduction in MAE of 1.93 %. The combined models produced a reduction in the dew point temperature MAE for only six of twelve prediction horizons with essentially no average decrease in MAE. However, the combined models showed a marked reduction in prediction anomalies for all twelve prediction horizons with an average reduction of 34.1 %. The reduction in prediction anomalies ranged from 4.6 % at the one-hour horizon to 60.5 % at the eleven-hour horizon.

A hybrid strategy of short term wind power prediction

Abstract Two different prediction methods are investigated for short term wind power prediction of a wind farm in this paper. The adopted strategies are individual artificial neural network (ANN) and hybrid strategy based on the physical and the statistical methods. The performance of two prediction methods is comprehensively compared. The calculated results show that the individual ANN prediction method can yield the prediction results quickly. The prediction accuracy is low and the root mean squared error (RMSE) is 10.67%. By contrast the hybrid prediction method operates costly and slowly. However, the prediction accuracy is high and the RMSE is 2.01%, less than 1/5 of that by individual ANN method. Meanwhile, it is found that the errors of the prediction have some relation with the wind speeds. The prediction errors are small when the wind speeds lower than 5 m/s or higher than 15 m/s. The reasons for such phenomena are also investigated.