Error Metrics Research Articles

COMPRESSIVE STRENGTH PREDICTION FOR SELF-COMPACTING CONCRETE USING ARTIFICIAL NEURAL NETWORKS

This study investigates the relationship between concrete constituent materials, slump values, density, strain and compressive strength. The database was set in the laboratory by performing a quality assurance test on the constituent materials to ascertain their suitability for concrete work. For material procurement and selection, the British Standard Department of Environment DOE technique of concrete trial mix designs was used. In this study, twelve concrete mixes were utilized. Two simplified models were developed using the feedback network Artificial Neural Networks architecture (ANNs) with R version 4.0.5 and R studio version 1.2.5033. In the two models, an independent layer comprising six nodes and a dependent layer comprising two nodes were taken. Having carried out the sensitivity analysis, the capacity of the developed equation was evaluated in terms of error metrics MSE and RMSE. The grading envelope for river sand showed that the graph fell perfectly into grading envelope zone 2 while other constituent materials tested were all confirmed suitable for concrete works. The results of compressive strength were obtained from varied water-cement ratios used for the concrete trial mix design which ranges between 0.45 to 0.6. The compressive strength results showed that the results increased consistently for each design mix. It is worthy of note that the strength of the mix fell between grades 15, 20, 25 and 35 respectively for both the 7-day and 28-day strength accordingly. Conversely, the ANN models predicted both 7-day and 28-day strength close to the laboratory value. It is concluded that this research has demonstrated economic viability which would enhance overall construction planning efficiency for future researchers, consultants and contractors in the building industries.

Enhanced SWAT calibration through intelligent range-based parameter optimization

Hydrological models are vital tools in environmental management. Weaknesses in model robustness for hydrological parameters transfer uncertainties to the model outputs. For streamflow, the optimized parameters are the primary source of uncertainty. A reliable calibration approach that reduces prediction uncertainty in model simulations is crucial for enhancing model robustness and reliability. The optimization of parameter ranges is a key aspect of parameter calibration, yet there is a lack of literature addressing the optimization of parameter ranges in hydrological models. In this paper, we introduce a parameter calibration strategy that applies a clustering technique, specifically the Self-Organizing Map (SM), to intelligently navigate the parameter space during the calibration of the Soil and Water Assessment Tool (SWAT) model for monthly streamflow simulation in the Baishan Basin, Jilin Province, China. We selected the representative algorithm, the Sequential Uncertainty Fitting version 2 (SUFI-2), from the commonly used SWAT Calibration and Uncertainty Programs for comparison. We developed three schemes: SUFI-2, SUFI-2-Narrowing Down (SUFI-2-ND), and SM. Multiple diagnostic error metrics were used to compare simulation accuracy and prediction uncertainty. Among all schemes, SM outperformed the others in describing watershed streamflow, particularly excelling in the simulation of spring snowmelt runoff (baseflow period). Additionally, the prediction uncertainty was effectively controlled, demonstrating the SM's adaptability and reliability in the interval optimization process. This provides managers with more credible prediction results, highlighting its potential as a valuable calibration tool in hydrological modeling.

Modelling Student Retention in Tutorial Classes with Uncertainty—A Bayesian Approach to Predicting Attendance-Based Retention

A Bayesian additive regression tree (BART) is a recent statistical method that blends ensemble learning with nonparametric regression. BART is constructed using a Bayesian approach, which provides the benefit of model-based prediction uncertainty, enhancing the reliability of predictions. This study proposes the development of a BART model with a binomial likelihood to predict the percentage of students retained in tutorial classes using attendance data sourced from a South African university database. The data consist of tutorial dates and encoded (anonymized) student numbers, which play a crucial role in deriving retention variables such as cohort age, active students, and retention rates. The proposed model is evaluated and benchmarked against the random forest regressor (RFR). The proposed BART model reported an average of 20% higher predictive performance compared to RFR across six error metrics, achieving an R-squared score of 0.9414. Furthermore, the study demonstrates the utility of the highest density interval (HDI) provided by the BART model, which can help in determining the best- and worst-case scenarios for student retention rate estimates. The significance of this study extends to multiple stakeholders within the educational sector. Educational institutions, administrators, and policymakers can benefit from this study by gaining insights into how future tutorship programme student retention rates can be predicted using predictive models. Furthermore, the foresight provided by the predicted student retention rates can aid in strategic resource allocation, facilitating more informed planning and budgeting for tutorship programmes.

A New Performance Metric to Evaluate Filter Feature Selection Methods in Text Classification

High dimensionality and sparsity are the primary issues in text classification. Using feature selection approaches, the most effective way to solve the problem is to select a subset of features. The most common and effective methods used for this process are filter techniques. Various performance metrics such as Micro-F1, Macro-F1, and Accuracy are used to evaluate the performance of filter methods used for feature selection on datasets  Such methods work depending on a classification algorithm. However, when selecting features in filter techniques, the information on the individual features is evaluated without considering the relationship between the features. In such an approach, the actual performance of the filter technique used in feature selection may not be determined. In such a case, it causes the existing methods to be insufficient in testing the validity of the proposed method. For this purpose, this study suggests a novel performance metric called Selection Error (SE) to determine the actual performance evaluation of filter techniques. The Selection Error metric allows us to analyze the information value of the selected features more accurately than existing methods without relying on a classifier. The feature selection performance of the filtering approaches was performed on six different datasets with both The Selection Error and traditional performance metrics. When the results are examined, it is seen that there is a strong relationship between the proposed performance metric and the classification performance metric results. The Selection Error aims to significantly contribute to the literature by demonstrating the success of filtering feature selection methods, regardless of classifier performance. 

The magnitude and frequency of detected precipitation determine the accuracy performance of precipitation data sets in the high mountains of Asia

The grid-based precipitation dataset is an important source for studying precipitation change in the high mountains of Asia due to where precipitation stations are sparse. It is essential to evaluate the accuracy of grid-based precipitation datasets in the high mountains of Asia before selecting an appropriate grid-based dataset. Therefore, this study comprehensively evaluated the precipitation errors of four commonly utilized precipitation datasets (multi-source weighted-ensemble precipitation (MSWEP), global precipitation climatology centre (GPCC), global precipitation measurement (GPM), and soil moisture to rain-advanced scatterometer (SM2RAIN-ASCAT)) in the high mountains of Asia from temporal and spatial perspectives. It then decomposed the precipitation errors to reveal their sources. The results showed that MSWEP, GPCC, GPM, and SM2RAIN-ASCAT overestimated precipitation amount and probability compared with station observations. Meanwhile, all precipitation data sets except MSWEP data underestimated precipitation in the dry season. In terms of the average values of the error metrics, GPCC performed the best. There was an evident annual periodicity in the error assessment metrics for the four precipitation data sets. Multiple linear regression analysis revealed that four precipitation-related factors (false alarm precipitation, missed amount of precipitation, precipitation detected presented, and precipitation detected event) explained the root mean square error values for four precipitation data sets, with precipitation detected presented having the largest weight. The root mean square error of each product exhibits periodic fluctuations with changes in precipitation quantity, attributed to the occurrences of precipitation detected presented and precipitation detected events. These findings provide useful reference information for correcting biases in precipitation data sets for high mountains of Asia.

Exogenous variable driven deep learning models for improved price forecasting of TOP crops in India

Accurately predicting agricultural commodity prices is crucial for India's economy. Traditional parametric models struggle with stringent assumptions, while machine learning (ML) approaches, though data-driven, lack automatic feature extraction. Deep learning (DL) models, with advanced feature extraction and predictive abilities, offer a promising solution. However, their application to agricultural price data ignored the exogenous factors. Hence, the study explored advanced versions of the well-known univariate models, NBEATSX and TransformerX. The research employed price data for essential crops like Tomato, Onion, and Potato (TOP) from major Indian markets and complemented it with corresponding weather data (precipitation and temperature). To provide a comprehensive analysis, the study also evaluated traditional statistical methods (ARIMAX and MLR) and a suite of ML algorithms (ANN, SVR, RFR, and XGBoost). The performance of these models was rigorously evaluated using error metrics like RMSE, MAE, sMAPE, MASE and QL. The findings were significant indicating DL models, particularly when augmented with exogenous variables, consistently outshone other methods with NBEATSX and TransformerX showing an average RMSE of 110.33 and 135.33, MAE of 60.08 and 74.92, sMAPE of 22.14 and 24.00, MASE of 1.02 and 1.32 and QL of 30.04 and 34.07, respectively. They exhibited lower error metrics, as compare to the statistical and ML models underscoring their effectiveness and potential in agricultural crop price forecasting. This study not only bridged a crucial research gap but also highlighted the robust potential of DL models in enhancing the accuracy of agricultural commodity price predictions in India.

Enhancing algal bloom forecasting: A novel framework for machine learning performance evaluation during periods of special temporal patterns

The evaluation of algal bloom forecasting models typically relies on error metrics that quantify the forecasting performance over the whole test set as a single number. Furthermore, the comparison with simple baseline methods is often omitted. To address this, we introduce a novel framework for Model performance Analysis and Visualization of time series forecasting (MAVts). MAVts incorporates novel algorithms for the automatic identification and visualization of time series periods of interest where the forecasting models are evaluated and compared with simple baseline methods. The application of MAVts on evaluating algal bloom forecasting models composed of sophisticated machine learning (ML) methods, reveals that in 85% of experiments a single error metric is not enough and only in 12.5% of experiments a ML model outperforms all baselines on all metrics and periods of interest. Thus, MAVts emerges as a valuable tool for analyzing and comparing ML models, advancing environmental management and protection.

An innovative moment balanced inference engine for predicting recycled concrete aggregate strength and minimizing mixture CO2 emissions

Recycled concrete aggregate (RCA) has an important role to play in reducing CO2 emissions and encouraging sustainable practices in the construction industry. This study introduces Optical Microscope Algorithm - Moment Balance Machine (OMA-MBM) as an innovative AI-based inference engine to improve the accuracy of RCA strength prediction. MBM improves the principles of Support Vector Machine (SVM) by considering moments to identify the optimal moment hyperplane. Backpropagation Neural Network (BPNN) is employed for weight assignment, and OMA is utilized for parameter optimization. The performance of the developed OMA-MBM was benchmarked against several machine learning models, including SVM, BPNN, PSO-SVM, and GWO-SVM, achieving the lowest error metrics with RMSE values of 9.711, 0.708, and 0.709 for compressive, flexural, and tensile strengths, respectively. It also scored the highest overall performance, represented by Reference Index of 1.000. Moreover, the model was successfully used to identify optimal RCA mixture designs in terms of achieving designated material strength requirements with the lowest CO2 emissions. The robustness of model predictions was further validated across diverse applications, i.e., Boston housing prices, vehicle fuel consumption, and building energy efficiency. These findings confirm the robustness and generalizability of OMA-MBM as an accurate predictive tool for development of sustainable construction.

A Comparative Study on Imputation Techniques: Introducing a Transformer Model for Robust and Efficient Handling of Missing EEG Amplitude Data.

In clinical datasets, missing data often occur due to various reasons including non-response, data corruption, and errors in data collection or processing. Such missing values can lead to biased statistical analyses, reduced statistical power, and potentially misleading findings, making effective imputation critical. Traditional imputation methods, such as Zero Imputation, Mean Imputation, and k-Nearest Neighbors (KNN) Imputation, attempt to address these gaps. However, these methods often fall short of accurately capturing the underlying data complexity, leading to oversimplified assumptions and errors in prediction. This study introduces a novel Imputation model employing transformer-based architectures to address these challenges. Notably, the model distinguishes between complete EEG signal amplitude data and incomplete data in two datasets: PhysioNet and CHB-MIT. By training exclusively on complete amplitude data, the TabTransformer accurately learns and predicts missing values, capturing intricate patterns and relationships inherent in EEG amplitude data. Evaluation using various error metrics and R2 score demonstrates significant enhancements over traditional methods such as Zero, Mean, and KNN imputation. The Proposed Model achieves impressive R2 scores of 0.993 for PhysioNet and 0.97 for CHB-MIT, highlighting its efficacy in handling complex clinical data patterns and improving dataset integrity. This underscores the transformative potential of transformer models in advancing the utility and reliability of clinical datasets.

A Prediction Model of Ionospheric Total Electron Content Based on Grid-Optimized Support Vector Regression

Evaluating and mitigating the adverse effects of the ionosphere on communication, navigation, and other services, as well as fully utilizing the ionosphere, have become increasingly prominent topics in the academic community. To quantify the dynamical changes and improve the prediction accuracy of the ionospheric Total Electron Content (TEC), we propose a prediction model based on grid-optimized Support Vector Regression (SVR). This modeling processes include three steps: (1) dividing the dataset for training, validation, and testing; (2) determining the hyperparameters C and g by the grid search method through cross-validation using training and validation data; and (3) testing the trained model using the test data. Taking the Gakona station as an example, we compared the proposed model with the International Reference Ionosphere (IRI) model and a TEC prediction model based on Statistical Machine Learning (SML). The performance of the models was evaluated using the metrics of mean absolute error (MAE) and root mean square error (RMSE). The specific results are as follows: the MAE of the CCIR, URSI, SML, and SVR models compared to the observations are 1.06 TECU, 1.41 TECU, 0.7 TECU, and 0.54 TECU, respectively; the RMSE are 1.36 TECU, 1.62 TECU, 0.92 TECU, and 0.68 TECU, respectively. These results indicate that the SVR model has the most minor prediction error and the highest accuracy for predicting TEC. This method also provides a new approach for predicting other ionospheric parameters.

Analyzing performance: Error-efficient, low-power recursive inexact multipliers for CNN applications

In present scenario, significant data manipulations can be espied in image processing and Convolutional Neural Networks (CNNs) applications. The quality performance of these applications relies ponderously on arithmetic operations. In this context, Approximate Computing frequently finds its application, where a certain margin of error is deemed acceptable, on prioritizing circuit speed over absolute precision and the common focus lies on optimizing the binary multiplier which is a crucial yet power-intensive component, so as to streamline the complexity of above mentioned applications. Furthermore, Inexact Multipliers (IMs) are developed by using recursive method, thereby purveying enhanced design and error metrics. This paper introduces higher order Recursive-Based Inexact Multipliers (RBIMs), which integrates the suggested 4-bit IMs, and exact multipliers along with the recursive method, aiming to provide a delicate balance between design performance metrics and error tolerance. For evaluating the proposed approach, most of the existing and recently suggested IMs operating at 8-bit and 16-bit, are designed in Verilog, coded in Python, synthesized and simulated using Cadence Register Transfer Language (RTL) Compiler, Google Colabs. Eventually, the simulation results bespeak a notable improvement in power consumption, with the 8-bit and 16-bit RBIMs by outperforming the previous designs up to 36.9 %, and 35.5 % respectively. Further, the efficacies of the RBIMs are validated through its usage in image processing and CNN applications.

A survey on node localization technologies in UWSNs: Potential solutions, recent advancements, and future directions

SummaryLocation‐based underwater communication applications such as strategic surveillance, disaster prevention, marine research, and mine detection have given the field of underwater wireless sensor networks (UWSN) a head start. Node localization is a prerequisite for accurate data collection, target monitoring, and network management in UWSNs. However, the unique characteristics of the underwater environment, such as signal attenuation, multipath propagation, and variable acoustic properties, pose a major challenge to effective node localization. Accurate sensor node location data is essential for successful underwater data collection, but difficult to achieve as the GPS system cannot be used in an underwater environment. In this paper, existing node localization techniques such as ALS, SLUM, MASL, SLMP, UDB, USP, etc., and recent advances such as the fusion of range‐based and range‐free techniques, the fusion of RSSI and AoA to improve localization accuracy by using directional information in addition to signal strength, and the use of optimization techniques such as PSO, COA, and WOA algorithms to improve the accuracy of the applied node localization algorithm, e.g., TP‐TSFLA, and challenges related to UWSN are discussed. Also, different localization algorithms that affect the accuracy of UWSN localization techniques have been evaluated and compared with NS2 in terms of localization error, localization coverage, energy consumption, and average communication cost metrics. In addition, this paper also provides an up‐to‐date investigation of localization techniques. Finally, the tools available for simulation are presented, followed by open research questions that need to be addressed in the localization of nodes.

Comparative analysis of actual evapotranspiration values estimated by METRIC model using LOCAL data and EEFlux for an irrigated area in Northern Sinaloa, Mexico

Landsat images combined with the METRIC model have been used in applications such as EEFlux to estimate actual evapotranspiration (ETa) in irrigated areas, with uncertainty as to whether the results are sufficiently accurate at local scales. This work compares temporally and spatially ETa estimates using Landsat imagery and the METRIC model with LOCAL measured weather data and EEFlux for an irrigated area in northern Sinaloa, Mexico, from 1995 to 2018. A regression analysis and error metrics such as coefficient of determination (R2), mean absolute error (MAE), root mean square error (RMSE) and slope were used to compare both models. The temporal analysis results confirm that ETa differences are closely related to crop growth and that the daily mean absolute error was 1.17 mm/day. Spatial analysis showed that when using only cropland pixels without non-cropland pixels in EEFlux, the R2 increases from 0.36 to 0.73, and the RMSE decreases from 2.52 to 1.98 mm/day. Therefore, EEFlux should be limited to cropland areas, and it is suggested to exclude Landsat images close to the planting and harvesting season to reduce the differences.

CWF: Consolidating Weak Features in High-quality Mesh Simplification

In mesh simplification, common requirements like accuracy, triangle quality, and feature alignment are often considered as a trade-off. Existing algorithms concentrate on just one or a few specific aspects of these requirements. For example, the well-known Quadric Error Metrics (QEM) approach [Garland and Heckbert 1997] prioritizes accuracy and can preserve strong feature lines/points as well, but falls short in ensuring high triangle quality and may degrade weak features that are not as distinctive as strong ones. In this paper, we propose a smooth functional that simultaneously considers all of these requirements. The functional comprises a normal anisotropy term and a Centroidal Voronoi Tessellation (CVT) [Du et al. 1999] energy term, with the variables being a set of movable points lying on the surface. The former inherits the spirit of QEM but operates in a continuous setting, while the latter encourages even point distribution, allowing various surface metrics. We further introduce a decaying weight to automatically balance the two terms. We selected 100 CAD models from the ABC dataset [Koch et al. 2019], along with 21 organic models, to compare the existing mesh simplification algorithms with ours. Experimental results reveal an important observation: the introduction of a decaying weight effectively reduces the conflict between the two terms and enables the alignment of weak features. This distinctive feature sets our approach apart from most existing mesh simplification methods and demonstrates significant potential in shape understanding. Please refer to the teaser figure for illustration.

A multi-task generative model for simultaneous post-contrast MR image synthesis and brainstem glioma segmentation

ObjectivesThis study aims to generate post-contrast MR images reducing the exposure of gadolinium-based contrast agents (GBCAs) for brainstem glioma (BSG) detection, simultaneously delineating the BSG lesion, and providing high-resolution contrast information. MethodsA retrospective cohort of 30 patients diagnosed with brainstem glioma was included. Multi-contrast images, including pre-contrast T1 weighted (pre-T1w), T2 weighted (T2w), arterial spin labeling (ASL) and post-contrast T1w images, were collected. A multi-task generative model was developed to synthesize post-contrast T1w images and simultaneously segment BSG masks from the multi-contrast inputs. Performance evaluation was conducted using peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and mean absolute error (MAE) metrics. A perceptual study was also undertaken to assess diagnostic quality. ResultsThe proposed model achieved SSIM of 0.86 ± 0.04, PSNR of 26.33 ± 0.05 and MAE of 57.20 ± 20.50 for post-contrast T1w image synthesis. Automated delineation of the BSG lesions achieved Dice similarity coefficient (DSC) score of 0.88 ± 0.27. ConclusionsThe proposed model can synthesize high-quality post-contrast T1w images and accurately segment the BSG region, yielding satisfactory DSC scores. Clinical relevance statementThe synthesized post-contrast MR image presented in this study has the potential to reduce the usage of gadolinium-based contrast agents, which may pose risks to patients. Moreover, the automated segmentation method proposed in this paper aids radiologists in accurately identifying the brainstem glioma lesion, facilitating the diagnostic process.

Transformer-Based Deep Learning Models for State of Charge and State of Health Estimation of Li-Ion Batteries: A Survey Study

The global transportation system’s need for electrification is driving research efforts to overcome the drawbacks of battery electric vehicles (BEVs). The accurate and reliable estimation of the states of charge (SOC) and health (SOH) of Li-Ion batteries (LIBs) is crucial for the widespread adoption of BEVs. Transformers, cutting-edge deep learning (DL) models, are demonstrating promising capabilities in addressing various sequence-processing problems. This manuscript presents a thorough survey study of previous research papers that introduced modifications in the development of Transformer-based architectures for the SOC and SOH estimation of LIBs. This study also highlights approximately 15 different real-world datasets that have been utilized for training and testing these models. A comparison is made between the architectures, addressing each state using the root mean square error (RMSE) and mean absolute error (MAE) metrics.

Prediction of Geometric Characteristics of Laser Cladding Layer Based on Least Squares Support Vector Regression and Crested Porcupine Optimization.

The morphology size of laser cladding is a crucial parameter that significantly impacts the quality and performance of the cladding layer. This study proposes a predictive model for the cladding morphology size based on the Least Squares Support Vector Regression (LSSVR) and the Crowned Porcupine Optimization (CPO) algorithm. Specifically, the proposed model takes three key parameters as inputs: laser power, scanning speed, and powder feeding rate, with the width and height of the cladding layer as outputs. To further enhance the predictive accuracy of the LSSVR model, a CPO-based optimization strategy is applied to adjust the penalty factor and kernel parameters. Consequently, the CPO-LSSVR model is established and evaluated against the LSSVR model and the Genetic Algorithm-optimized Backpropagation Neural Network (GA-BP) model in terms of relative error metrics. The experimental results demonstrate that the CPO-LSSVR model can achieve a significantly improved relative error of no more than 2.5%, indicating a substantial enhancement in predictive accuracy compared to other methods and showcasing its superior predictive performance. The high accuracy of the CPO-LSSVR model can effectively guide the selection of laser cladding process parameters and thereby enhance the quality and efficiency of the cladding process.

Quality-preserving multilevel mesh generation for building models

ABSTRACT Three-dimensional (3D) high-fidelity building models are crucial for digital twin cities. Multi-scale modeling is crucial for accurately capturing the complexity of urban environments; however, existing methods often lack the flexibility required to satisfy the varied demands of diverse applications. In this paper, we propose a framework to continuously simplify 3D building models, it dynamically adjusts mesh accuracy to meet diverse needs, enabling seamless transitions from high-poly to medium and low-poly meshes. Our method contains three main stages: First, in the mesh repair stage, we introduce a novel mesh repair framework based on an advanced Marching Cubes 33 algorithm to ensure a watertight and clean topology mesh. Then, for generating medium-poly meshes, we employ a feature-preserving mesh simplification algorithm that combines curvature and area-weighted quadratic error metrics to achieve a high-fidelity simplified mesh. For generating low-poly meshes, we use the concept of carving visual hull to generate low-poly mesh with high visual similarity. Our method has been evaluated on building subsets of the Free3D and Thingi10 K datasets, as well as on various building models generated from different observation scales, and its superiority has been validated through extensive comparisons with state-of-the-art techniques.

A novel optimized hybrid machine learning model to enhance the prediction accuracy of hourly building energy consumption

ABSTRACT Building energy consumption (BEC) prediction is crucial in efficient energy management. This paper proposes an optimized hybrid prediction model that combines a support vector regression (SVR), a newly evolved coati optimization algorithm (COA), and a recursive feature elimination with cross-validation (RFECV) implemented on an hourly new dataset. The SVR is selected based on the experimentations conducted in this work that outperform other models. The COA is used for optimizing the hyperparameters of SVR, and RFECV is used to optimize the dataset. The SVR COA performs better than the Harris hawk optimization and Gray wolf optimization. Later, the optimized SVR is implemented on the optimized dataset, showing better accuracy and faster prediction compared with the default SVR, SVR with feature elimination, and optimized SVR models. The error metrics MAE, MAPE, RMSE, and R2 are used for the model evaluation. The testing accuracy improved by 12.34%, 10.52%, 17.02%, and 0.09%, respectively, compared to the default model.

Precise single step and multistep short-term photovoltaic parameters forecasting based on reduced deep convolutional stack autoencoder and minimum variance multikernel random vector functional network

Accurate prediction of photovoltaic (PV) power can significantly alleviate energy crises. However, the inherent randomness and intermittency of PV power pose challenges to the stability and safety of PV-penetrated grid systems. To address this, we have developed a novel hybrid model: a reduced deep convolutional stack autoencoder with a minimum variance multikernel random vector functional link network (RDCSAE-MVMRVFLN). This model enhances grid efficacy and safety. We extract the most informative band-limited intrinsic mode functions (BLIMFs) of highly nonlinear and nonstationary solar energy parameters using an entropy, kurtosis, and correlation coefficient-based information-oriented variational mode decomposition (IOVMD). These efficient BLIMFs are concatenated and input into the RDCSAE for rich, abstract, and discriminative representation computation. A less computationally complex MVMKRVFLN regression method, incorporating these refined representations, is proposed for superior prediction accuracy and reduced computational complexity. Our method shows exceptional performance in predicting solar temperature, irradiation, and power for multi-horizon forecasts with minimal error metrics (correlation coefficients of 0.999±0.001, 0.992±0.001, 0.986±0.02 and 0.978±0.02, and RMSE of 0.016±0.001, 0.024±0.001, 0.034±0.001 and 0.045±0.001 for the interval of 10 minutes, 30 minutes, 1 hour and 3 hours respectively) in both single-step and multi-step forecasting compared to conventional methods. The RDCSAE-MVMKRVFLN model is implemented on a high-speed Xilinx Virtex-5 FPGA embedded processor to validate its simplicity, robustness, and practicability. Additionally, we examine the prediction performance using real-time data from a 1 MW solar farm in Odisha, India, demonstrating the model’s effectiveness and superiority.