Uncorrelated Data Research Articles

Top-[formula omitted] discriminative feature selection with uncorrelated and [formula omitted]-norm equation constraints

Supervised feature selection (FS) as an interpretable dimensionality reduction technique has received increasing attention, where linear discriminative analysis (LDA)-based method can select informative features discriminatively and obtain promising performance. When original data has more features than samples, however, LDA-based method generally encounters degradation since the appearance of irreversible scatter matrix. This situation is called the small sample size (SSS) problem. To overcome it and enhance the discriminant power of selected feature subsets, in this paper, we design an elegant LDA-based FS model referred to as Top-k Discriminative FS (TDFS), which is constructed by seamlessly integrating the ℓ2,0-norm equation constraint into uncorrelated LDA model. More concretely, the ℓ2,0-norm equation constraint can explicitly characterize the number of selective features k to ensure the sparsity of projected matrix and select top features. The uncorrelated LDA model aims to improve discriminative ability based on uncorrelated data in projected subspace. Given the formidable nature of solving this non-convex model, a novel optimization algorithm is further developed and the SSS problem can be efficaciously addressed during the optimization process. We first decompose projection matrix into a discrete selection matrix and its corresponding nonzero projection matrix, then concurrently optimize above two matrices by employing a column-by-column update scheme, during which the reversibility of scatter matrix in selective feature subspace can be easily guaranteed to solve SSS problem. The extensive experiments on four synthetic data sets and eight real-world data sets show that the proposed method outperforms eight competitors validated by three classifiers. Moreover, although the theoretical analysis proves that our algorithm has quartic time complexity on the number of selected features k, the running time experiments verify that TDFS is still efficient and applicable in scenarios where only a small number of features need to be selected. From above perspectives, our algorithm shows desirable performance to achieve discriminative FS.

Monitoring of high-dimensional and high-frequency data streams: A nonparametric approach

ABSTRACT High-dimensional and high-frequency data streams (HHDS) have become widely available with the prevalence of data-acquisition technologies. Because HHDS have the characteristics of high dimensionality and temporal correlation, traditional statistical process control (SPC) techniques cannot be used directly to address the problem of monitoring HHDS. Most extant monitoring methods commonly assume that the process observations are independent or can be described by some parametric models and face the challenge posed by high-dimensional data. Little research has been conducted on the robust monitoring of HHDS that can simultaneously accommodate high dimensions and high frequencies. Therefore, in this paper, we propose a novel nonparametric approach for monitoring HHDS, based on data decorrelation, dimension reduction, and SPC. Specifically, process observations are first sequentially decorrelated and standardized using Cholesky decomposition, and then the k -nearest neighbors classification algorithm is applied to transform uncorrelated high-dimensional data into one-dimensional data. Finally, the traditional cumulative-sum procedure is used for online monitoring, based on an empirical log-likelihood ratio test. Numerical studies have shown that our monitoring scheme is sufficiently reliable and efficient for the online monitoring of HHDS. An illustrative example about the Dow Jones 30 industrial stock prices is presented to further validate the proposed approach.

Breaking the baryon-dark matter degeneracy in a model-independent way through the Sunyaev-Zeldovich effect

Context. In cosmological fits, it is common to fix the baryon density ωb via the cosmic microwave background. We here constrain ωb by means of a model-independent interpolation of the acoustic parameter from correlated baryonic acoustic oscillations. Aims. The proposed technique is used to alleviate the degeneracy between baryonic and dark matter abundances. Methods. We propose a model-independent Bézier parametric interpolation and applied it to intermediate-redshift data. We first interpolated the observational Hubble data to extract cosmic bounds over the (reduced) Hubble constant h0 and interpolated the angular diameter distances, D(z), of the galaxy clusters, inferred from the Sunyaev-Zeldovich effect, to constrain the spatial curvature, Ωk. Through the Hubble points and D(z) determined in this way, we interpolated uncorrelated data of baryonic acoustic oscillations bounding the baryon ωb and total matter ωm densities, reinforcing the constraints on h0 and Ωk with the same technique. Finally, to remove the matter sector degeneracy, we obtained ωb by interpolating the acoustic parameter from correlated baryonic acoustic oscillations. Results. Monte Carlo Markov chain simulations agree at 1σ confidence level with the flat ΛCDM model and are roughly suitable at 1σ with its nonflat extension, while the Hubble constant appears in tension up to the 2σ confidence levels. Conclusions. Our method excludes very small extensions of the standard cosmological model, and on the Hubble tension side, seems to match local constraints slightly.

Reprint of: Robust inference on correlation under general heterogeneity

Considerable evidence in past research shows size distortion in standard tests for zero autocorrelation or zero cross-correlation when time series are not independent identically distributed random variables, pointing to the need for more robust procedures. Recent tests for serial correlation and cross-correlation in Dalla, Giraitis, and Phillips (2022) provide a more robust approach, allowing for heteroskedasticity and dependence in uncorrelated data under restrictions that require a smooth, slowly-evolving deterministic heteroskedasticity process. The present work removes those restrictions and validates the robust testing methodology for a wider class of innovations and regression residuals allowing for heteroscedastic uncorrelated and non-stationary data settings. The updated analysis given here enables more extensive use of the methodology in practical applications. Monte Carlo experiments confirm excellent finite sample performance of the robust test procedures even for extremely complex white noise processes. The empirical examples show that use of robust testing methods can materially reduce spurious evidence of correlations found by standard testing procedures.

Improving Short-Term Traffic Flow Prediction using Grey Relational Analysis for Data Filtering and Stacked LSTM Modeling

Traffic flow prediction is one of the critical measures to alleviate traffic congestion. Currently, traffic flow prediction research has made some achievements, but there are still some deficiencies. In order to solve the problems of low prediction accuracy, poor real-time performance, and high data dimensions. This paper proposes a new traffic flow prediction method that employs Grey Relation Analysis (GRA) to detect the correlation between detection points, remove insignificant or uncorrelated traffic flow data points, and hence reduce the data dimensionality of the prediction model. Multiple Long Short-Term Memory (LSTM) models are then stacked to establish the traffic flow prediction model, considering that traffic flow is affected by multi-dimensional spatiotemporal factors, incorporating vehicle speed, occupancy, and traffic volume as inputs. We conducted experiments on real datasets, and the results showed that our GRA-SLSTM model improved prediction accuracy by 3.6% compared to other models, while reducing model prediction time by 27.33%. The proposed model’s generalization ability is validated by predicting other detection points, which provides significant references for traffic flow prediction research and practical applications.

Generalised Least-Squares Fitting with Procedures for Uncorrelated Data of Constant Variance

Generalised Least-Squares Fitting with Procedures for Uncorrelated Data of Constant Variance

Stochastic Method for Generating Residential Household Energy Models of Varying Income Level and Climate Zone for Testing Energy Fairness

Abstract Recent developments in residential energy management necessitate testing of whether energy technology and policy fairly impact households of various income levels and locations. Considering income level is particularly important because increased utility bills put low-income households at higher risk of economic and health issues. This paper proposes a novel method for generating diverse household energy models of varying income levels and climate zones, which can be used to test the fairness impacts of residential energy developments. Models are stochastically generated using probability distributions based on data from national surveys. Included in the model are constant and time-variable features. Models capture the randomness inherent in the residential sector while still following realistic patterns of each income-climate group regarding building structure, appliance stock, and occupant behavior. Models were validated by comparing when, how, and how much energy is consumed by simulated versus real-life surveyed households. A total of 200 household models were simulated for the validation process, representing four climate zones and five income levels. Simulated energy consumption was plotted against survey data from the same income level, climate zone, and income-climate combination. Through correlation analysis and null hypothesis testing, it was determined that there is no statistically significant difference between simulated and surveyed energy consumption. For all cases considered, the correlation of the data is highly statistically significant. When validating all cases for income-climate classifications and individual climate zones, there was less than a 0.05% chance of uncorrelated data exhibiting such high correlation coefficients (r2), which ranged from 0.862 to 0.998. When validating models of each income level, this chance ranged from less than 0.05% to 0.194% with r2 values ranging from 0.816 to 0.984. A linear trendline was fitted to the data, and a null hypothesis test was performed to check if the slope statistically differed from 1. All cases tested resulted in a P value greater than 0.05 which, for a 95% confidence level, indicates that no significant difference can be determined. Because the plot of simulated versus survey data was very highly correlated and exhibited a slope statistically indistinguishable from 1, simulated household models were determined to represent real-life households with sufficient accuracy.

Investigating the impact of biomimetic pinna shape variations on clutter echoes received from natural environments

Bat species navigating dense vegetation based on biosonar have to obtain the necessary sensory information from “clutter echoes”, i.e., echoes that are superpositions of contributions from many reflecting facets (e.g., leaves) and hence have highly unpredictable waveforms. Prior results have suggested that pinna motions could aid in direction-finding tasks based on deterministic echo patterns. This raises the question whether varying pinna shapes could also have a function significance for challenging biosonar tasks performed on clutter echoes. As a first, task-independent step to test this hypothesis it has been investigated whether different pinna shapes have a consistent effect on clutter echoes despite the random nature of these signals. This was accomplished using a dedicated laboratory setup that produced large amounts of uncorrelated clutter echo data by agitating an artificial foliage with fans between echoes. Deep learning methods were then applied to identify the pinna shape that received a given clutter echo. A two-dimensional convolutional neural network operating on spectrograms achieved 90% validation accuracy in this task. This finding demonstrates that even small pinna deformations can impart consistent effects on the clutter echoes. Ongoing research is directed at analyzing the nature of the signal properties that the successful classifications were based on.

Development of an error correction method using perfect binary arrays

The research focuses on an innovative error correction method that uses perfect binary arrays (PBAs), a powerful mathematical tool with unique properties that make it ideal for error correction. The research is aimed at studying the impact of uncorrelated mixed-type errors in the data exchange path, which allows using it in smart technologies with limited computing capabilities. The effectiveness of the approach is confirmed by simulation and comparison with other error correction methods. In order to further study the structural, cross-correlation and distance properties of orthogonal two-dimensional codes and the correcting capabilities of the proposed method, an information technology system for data transmission based on an equivalent class of perfect binary arrays has been developed. The proposed model evaluates the performance of the error correction code based on perfect binary arrays under various conditions, including correlated and uncorrelated interference and data exchange paths. A generator of PBA of equivalent classes has been built. An experimental evaluation of the correcting ability of the proposed two-dimensional codes was carried out by simulating various pre-code situations, including packet and random errors, for the cases of correlated and uncorrelated interference. Using a graphical interface, users will be able to enter the number and type of errors, determine whether they are random or packet errors, manually or automatically, move errors through the data packet, and view intermediate results. Thus, the complex nature of this study can be positioned as a promising approach and a reliable choice in the field of error correction

Principal Component Analysis of Selective Bands for Digital Satellite Imagery Classification

Principal Component Analysis technique was applied to correlated multispectral data to convert it into uncorrelated data for the purpose of benefiting from its use in the automatic classification process.This project aims to classify multispectral satellite images using selective thematic Mapper (TM) spectral bands, and principal component (PC) images. The supervised classification method with maximum likelihood is adopted to perform the classification process.This technique was implemented using banding images and PC’s images. The results showed that the classification accuracy of the three spectral bands with the highest contrast was 86.9%, compared to the classification accuracy of the first three PC’s (89.08%). This value represents the highest classification accuracy obtained among the classification accuracy values for the spectral bands or PC images alone or in combination.The results show the advantages of feature selection in the PC’s, for every value of n-components, the 1stprincipal is the best choice. Moreover at low values of n-PC’s contain more information about the discriminability of classes than for any combination of n-original spectral channels.

BSEIFFS: Blockchain-secured edge-intelligent forest fire surveillance

This work aims to address two major issues in forest fire management: accurate real-time forecast and end-to-end secured data delivery between validated entities. To do so, we introduce BSEIFFS, a secured and intelligent forest fire prediction architecture realized through the convergence of Blockchain technology, mist computing, edge computing and machine learning paradigms. The proposed architecture renders Blockchain-backed security and Support Vector Regression based wildfire prediction at the edge by offloading the execution overheads to containerized cloud applications. Two algorithms namely Uncorrelated Data Transmission (UDT) algorithm and Blockchain based Audit Trail validation (BAT) algorithm are proposed to reduce number of repetitive packet transmissions and rendering a lightweight Blockchain based identity validation scheme at the edge respectively. The prediction model is trained using real-world meteorological data and tested on a steel manufacturing facility where controlled combustion and quenching processes closely resemble fire and rain scenarios. Apart from the Google Cloud Platform, tools such as IBM NodeRed, Siemens IBA PDA softwares are used to implement the BSEIFFS architecture. Extensive experimentation and comparative evaluation are performed with relevant baseline approaches. BSEIFFS achieves a prediction accuracy of 98.61%, ROC AUC score of 0.9668 with 0.1127 as false positive rate. The BSEIFFS SVR predictor with rbf kernel (MAE = 1.840, RMSE = 5.728) performs better than similar baseline approaches. The proposed UDT algorithm causes 82.02% reduction in packet transmissions through the network. BSEIFFS also achieves atleast 44.84% lesser energy consumption, 11.91% higher throughput and 71.91% lesser roundtrip time in comparison with baseline approaches. BSEIFFS can massively extend the benefits of the imminent B5G/6G technologies to perform fine-grained and faster computations.

A Semi-Supervised Approach for Improving Generalization in Non-Intrusive Load Monitoring.

Non-intrusive load monitoring (NILM) considers different approaches for disaggregating energy consumption in residential, tertiary, and industrial buildings to enable smart grid services. The main feature of NILM is that it can break down the bulk electricity demand, as recorded by conventional smart meters, into the consumption of individual appliances without the need for additional meters or sensors. Furthermore, NILM can identify when an appliance is in use and estimate its real-time consumption based on its unique consumption patterns. However, NILM is based on machine learning methods and its performance is dependent on the quality of the training data for each appliance. Therefore, a common problem with NILM systems is that they may not generalize well to new environments where the appliances are unknown, which hinders their widespread adoption and more significant contributions to emerging smart grid services. The main goal of the presented research is to apply a domain adversarial neural network (DANN) approach to improve the generalization of NILM systems. The proposed semi-supervised algorithm utilizes both labeled and unlabeled data and was tested on data from publicly available REDD and UK-DALE datasets. The results show a 3% improvement in generalization performance on highly uncorrelated data, indicating the potential for real-world applications.

An IoT-based resource utilization framework using data fusion for smart environments

Nowadays, many communities are emerging towards smart environments, requiring the communication and collaboration of diverse Internet-of-Things (IoT) devices. A smart environment exploits the use of IoT technology to share and process data among such devices for a better living. However, this comes with additional costs, such as the exponential growth of IoT devices, the heterogeneity of IoT use cases, and the new complex features encountered by IoT data, which complicate their processing and analysis using the traditional techniques. This causes a dramatic performance degradation of the used processing resources, which directly affects the overall efficiency and performance of IoT-based systems. Although different studies have presented resource utilization approaches for IoT systems, but they were not evaluated from different resource utilization perspectives. Besides, no efforts have been directed to investigate their effectiveness to process the unprecedented IoT data features that inevitably impact the accuracy and efficiency of resource utilization. In this paper, the Triple Phases Resource Utilized Data Fusion (TPRUDF) framework is proposed as the first IoT-based cost-aware resource utilization using data fusion. It exclusively considers different IoT data features by employing three phases of data fusion: (1) data in – data out, (2) data in – feature out, and (3) feature in – decision out. TPRUDF fuses the raw IoT data by maintaining the complex IoT data features, independent of the IoT domain or the computing model, using the spatiotemporal data fusion (STDF) IoT-based data fusion approach. TPRUDF then fuses the uncorrelated data features via the Principal Component Analysis. Finally, it employs two different resource utilization techniques: (1) Genetic Algorithms and (2) Particle Swarm Optimization, fusing their results using the voting logic fusion technique. A public edge-computing simulator is used to evaluate TPRUDF via three real-world smart cities datasets. The experimental results of the proposed TPRUDF framework indicate that it: (1) achieves an average accuracy level of resource utilization equal to 91%, (2) increases the resource utilization throughput by an average of 40% and eventually minimizes the processing delay, (3) boosts the resource utilization availability by 60%, and (4) decreases the energy consumption by 35%.

Uncertainty-aware automated machine learning toolbox

Abstract Measurement data can be considered complete only with an associated measurement uncertainty to express knowledge about the spread of values reasonably attributed to the measurand. Measurement uncertainty also allows to assess the comparability and the reliability of measurement results as well as to evaluate decisions based on the measurement result. Artificial Intelligence (AI) methods and especially Machine Learning (ML) are often based on measurements, but so far, uncertainty is widely neglected in this field. We propose to apply uncertainty propagation in ML to allow estimating the uncertainty of ML results and, furthermore, an optimization of ML methods to minimize this uncertainty. Here, we present an extension of a previously published automated ML toolbox (AMLT), which performs feature extraction, feature selection and classification in an automated way without any expert knowledge. To this end, we propose to apply the principles described in the “Guide to the Expression of Uncertainty in Measurement” (GUM) and its supplements to carry out uncertainty propagation for every step in the AMLT. In previous publications we have presented the uncertainty propagation for some of the feature extraction methods in the AMLT. In this contribution, we add some more elements to this concept by also including statistical moments as a feature extraction method, add uncertainty propagation to the feature selection methods and extend it to also include the classification method, linear discriminant analysis combined with Mahalanobis distance. For these methods, analytical approaches for uncertainty propagation are derived in detail, and the uncertainty propagation for the other feature extraction and selection methods are briefly revisited. Finally, the use the uncertainty-aware AMLT is demonstrated for a data set consisting of uncorrelated measurement data and associated uncertainties.

Quasi-simultaneous observations of radio and X-ray variability in three radio-quiet Seyfert galaxies

ABSTRACT Radio variability in some radio-quiet (RQ) active galactic nuclei suggests emission from regions close to the central engine, possibly the outer accretion disc corona. If the origins of the radio and the X-ray emission are physically related, their emission may be temporarily correlated, possibly with some time delays. We present the results of quasi-simultaneous radio and X-ray monitoring of three RQ Seyfert galaxies, Mrk 110, Mrk 766, and NGC 4593, carried out with the Very Large Array at 8.5 GHz over a period of about 300 d, and with the Rossi X-ray Timing Explorer at 2–10 keV over a period of about 2000 d. The radio core variability is likely detected in the highest resolution (A configuration) observations of Mrk 110 and NGC 4593, with a fractional variability amplitude of 6.3 per cent and 9.5 per cent, respectively. A cross-correlation analysis suggests an apparently strong (Pearson r = −0.89) and highly significant correlation (p = 1 × 10−6) in Mrk 110, with the radio lagging the X-ray by 56 d. However, a further analysis of the r values distribution for physically unrelated long time delays reveals that this correlation is not significant. This occurs since the Pearson correlation assumes white noise, while both the X-ray and the radio light curves follow red noise, which dramatically increases the chance, by a factor of ∼103, to get extremely high r values in uncorrelated data sets. A significantly longer radio monitoring with a higher sampling rate, preferably with a high-resolution fixed radio array, is required in order to reliably detect a delay.

Corrupted bifractal features in finite uncorrelated power-law distributed data

Multifractal Detrended Fluctuation Analysis stands out as one of the most reliable methods for unveiling multifractal properties, specially when real-world time series are under analysis. However, little is known about how several aspects, like artefacts during the data acquisition process, affect its results. In this work we have numerically investigated the performance of Multifractal Detrended Fluctuation Analysis applied to synthetic finite uncorrelated data following a power-law distribution in the presence of additive noise, and periodic and randomly-placed outliers. We have found that, on one hand, spurious multifractality is observed as a result of data finiteness, while additive noise leads to an underestimation of the exponents hq for q<0 even for low noise levels. On the other hand, additive periodic and randomly-located outliers result in a corrupted inverse multifractality around q=0. Moreover, the presence of randomly-placed outliers corrupts the entire multifractal spectrum, in a way proportional to their density. As an application, the multifractal properties of the time intervals between successive aircraft landings at three major European airports are investigated.

An investigation of penalization and data augmentation to improve convergence of generalized estimating equations for clustered binary outcomes

BackgroundIn binary logistic regression data are ‘separable’ if there exists a linear combination of explanatory variables which perfectly predicts the observed outcome, leading to non-existence of some of the maximum likelihood coefficient estimates. A popular solution to obtain finite estimates even with separable data is Firth’s logistic regression (FL), which was originally proposed to reduce the bias in coefficient estimates. The question of convergence becomes more involved when analyzing clustered data as frequently encountered in clinical research, e.g. data collected in several study centers or when individuals contribute multiple observations, using marginal logistic regression models fitted by generalized estimating equations (GEE). From our experience we suspect that separable data are a sufficient, but not a necessary condition for non-convergence of GEE. Thus, we expect that generalizations of approaches that can handle separable uncorrelated data may reduce but not fully remove the non-convergence issues of GEE.MethodsWe investigate one recently proposed and two new extensions of FL to GEE. With ‘penalized GEE’ the GEE are treated as score equations, i.e. as derivatives of a log-likelihood set to zero, which are then modified as in FL. We introduce two approaches motivated by the equivalence of FL and maximum likelihood estimation with iteratively augmented data. Specifically, we consider fully iterated and single-step versions of this ‘augmented GEE’ approach. We compare the three approaches with respect to convergence behavior, practical applicability and performance using simulated data and a real data example.ResultsOur simulations indicate that all three extensions of FL to GEE substantially improve convergence compared to ordinary GEE, while showing a similar or even better performance in terms of accuracy of coefficient estimates and predictions. Penalized GEE often slightly outperforms the augmented GEE approaches, but this comes at the cost of a higher burden of implementation.ConclusionsWhen fitting marginal logistic regression models using GEE on sparse data we recommend to apply penalized GEE if one has access to a suitable software implementation and single-step augmented GEE otherwise.

Simultaneous mnemonic and predictive representations in the auditory cortex

Recent studies have shown that stimulus history can be decoded via the use of broadband sensory impulses to reactivate mnemonic representations.1-4. However, memories of previous stimuli can also be used to form sensory predictions about upcoming stimuli.5,6 Predictive mechanisms allow the brain to create a probable model of the outside world, which can be updated when errors are detected between the model predictions and external inputs. 7-10 Direct recordings in the auditory cortex of awake mice established neural mechanisms for how encoding mechanisms might handle working memory and predictive processes without "overwriting" recent sensory events in instances where predictive mechanisms are triggered by oddballs within a sequence.11 However, it remains unclear whether mnemonic and predictive information can be decoded from cortical activity simultaneously during passive, implicit sequence processing, even in anesthetized models. Here, we recorded neural activity elicited by repeated stimulus sequences using electrocorticography (ECoG) in the auditory cortex of anesthetized rats, where events within the sequence (referred to henceforth as "vowels," for simplicity) were occasionally replaced with a broadband noise burst or omitted entirely. We show that both stimulus history and predicted stimuli can be decoded from neural responses to broadband impulses, at overlapping latencies but based on independent and uncorrelated data features. We also demonstrate that predictive representations are dynamically updated over the course of stimulation.

Enhancing Big Data Feature Selection Using a Hybrid Correlation-Based Feature Selection

This study proposes an alternate data extraction method that combines three well-known feature selection methods for handling large and problematic datasets: the correlation-based feature selection (CFS), best first search (BFS), and dominance-based rough set approach (DRSA) methods. This study aims to enhance the classifier’s performance in decision analysis by eliminating uncorrelated and inconsistent data values. The proposed method, named CFS-DRSA, comprises several phases executed in sequence, with the main phases incorporating two crucial feature extraction tasks. Data reduction is first, which implements a CFS method with a BFS algorithm. Secondly, a data selection process applies a DRSA to generate the optimized dataset. Therefore, this study aims to solve the computational time complexity and increase the classification accuracy. Several datasets with various characteristics and volumes were used in the experimental process to evaluate the proposed method’s credibility. The method’s performance was validated using standard evaluation measures and benchmarked with other established methods such as deep learning (DL). Overall, the proposed work proved that it could assist the classifier in returning a significant result, with an accuracy rate of 82.1% for the neural network (NN) classifier, compared to the support vector machine (SVM), which returned 66.5% and 49.96% for DL. The one-way analysis of variance (ANOVA) statistical result indicates that the proposed method is an alternative extraction tool for those with difficulties acquiring expensive big data analysis tools and those who are new to the data analysis field.

Fuzzy clustering-based neural networks modelling reinforced with the aid of support vectors-based clustering and regularization technique

In recent years, classical fuzzy clustering-based neural networks (FCNNs) have been successfully applied to regression tasks. The determination of the parameters such as cluster centers of the existing hard c-means (HCM) or fuzzy c-means (FCM), leads to the performance deterioration of the model because of the sensitivity of HCM or FCM to noise and outliers. Moreover, there are also several factors for over-fitting and degradation of the robustness of the ensuing model. To solve such problems, two improved clustering techniques and L2 norm-regularization are considered in the proposed robust fuzzy clustering-based neural networks (RFCNNs) modeling. SVs-based hard c-means (SVs-based HCM) and SVs-based fuzzy c-means (SVs-based FCM) designed with support vectors (SVs) can reduce the interference of uncorrelated data, including noise and outliers, thereby enhancing the main data characteristics effectively, as well as leading to the construction on the improved network model. L2 norm-regularization can be used to alleviate the degradation of robustness caused by overfitting. In terms of improving the performance of the model through SVs-based HCM or SVs-based FCM, as well as robustness completed through L2 norm-regularization, the superiority of RFCNNs was verified by experimenting with synthetic data and publicly available data from machine learning datasets.