SVR Model Research Articles

Stock Prediction using LSTM model

With the development of the times, investors are increasingly in demand for stock price forecasting. However, stock price fluctuations are full of uncertainty, making traditional machine learning algorithms more erroneous in long-term forecasting. Based on the LSTM model, this paper uses Tushare to obtain the historical price of stocks, and the optimal structure and best training parameters of the LSTM model in stock price prediction are determined experimentally. The prediction accuracy of the LSTM model was evaluated by MAE, and the best result was 69.15, which achieved accurate prediction of stock prices. Compared with the traditional SVR model and the ARMA model, the prediction results of LSTM are more in line with the actual value, and the prediction accuracy of the algorithm is higher.

A Model Selection Approach for Time Series Forecasting: Incorporating Google Trends Data in Australian Macro Indicators.

This study examined whether the behaviour of Internet search users obtained from Google Trends contributes to the forecasting of two Australian macroeconomic indicators: monthly unemployment rate and monthly number of short-term visitors. We assessed the performance of traditional time series linear regression (SARIMA) against a widely used machine learning technique (support vector regression) and a deep learning technique (convolutional neural network) in forecasting both indicators across different data settings. Our study focused on the out-of-sample forecasting performance of the SARIMA, SVR, and CNN models and forecasting the two Australian indicators. We adopted a multi-step approach to compare the performance of the models built over different forecasting horizons and assessed the impact of incorporating Google Trends data in the modelling process. Our approach supports a data-driven framework, which reduces the number of features prior to selecting the best-performing model. The experiments showed that incorporating Internet search data in the forecasting models improved the forecasting accuracy and that the results were dependent on the forecasting horizon, as well as the technique. To the best of our knowledge, this study is the first to assess the usefulness of Google search data in the context of these two economic variables. An extensive comparison of the performance of traditional and machine learning techniques on different data settings was conducted to enable the selection of an efficient model, including the forecasting technique, horizon, and modelling features.

Estimation of load capacity of driven piles by COA–MLP and COA–SVR models: a comparative study

Estimation of load capacity of driven piles by COA–MLP and COA–SVR models: a comparative study

Urban Flood Risk Assessment through the Integration of Natural and Human Resilience Based on Machine Learning Models

Flood risk assessment and mapping are considered essential tools for the improvement of flood management. This research aims to construct a more comprehensive flood assessment framework by emphasizing factors related to human resilience and integrating them with meteorological and geographical factors. Moreover, two ensemble learning models, namely voting and stacking, which utilize heterogeneous learners, were employed in this study, and their prediction performance was compared with that of traditional machine learning models, including support vector machine, random forest, multilayer perceptron, and gradient boosting decision tree. The six models were trained and tested using a sample database constructed from historical flood events in Hefei, China. The results demonstrated the following findings: (1) the RF model exhibited the highest accuracy, while the SVR model underestimated the extent of extremely high-risk areas. The stacking model underestimated the extent of very-high-risk areas. It should be noted that the prediction results of ensemble learning methods may not be superior to those of the base models upon which they are built. (2) The predicted high-risk and very-high-risk areas within the study area are predominantly clustered in low-lying regions along the rivers, aligning with the distribution of hazardous areas observed in historical inundation events. (3) It is worth noting that the factor of distance to pumping stations has the second most significant driving influence after the DEM (Digital Elevation Model). This underscores the importance of considering human resilience factors. This study expands the empirical evidence for the ability of machine learning methods to be employed in flood risk assessment and deepens our understanding of the potential mechanisms of human resilience in influencing urban flood risk.

Machine-Learning-Based Electric Power Forecasting

The regional demand for electric power is influenced by a variety of factors, such as fluctuations in business cycles, dynamic linkages among regional development, and climate change. The valid quantification of the impacts of these factors on the demand for electric power poses significant challenges. Existing methods often fall short of capturing the inherent complexities. This paper addresses these limitations by proposing a framework, which integrates machine-learning techniques into regional electricity demand forecasting. Regional electricity generation firms could then leverage the power of machine learning and improve the accuracy and robustness of electric power forecasting. In this paper, we conduct extensive numerical experiments using an actual dataset from a large utility firm and other public data sources. The analysis indicates that the support vector regression model (the SVR model) has high accuracy in predicting the demand. The results show that socio-economic development is the major driver of growth in electricity demand, while weather variability is a key contributor to the seasonal fluctuations in electricity use. Furthermore, linkages among regional development and the status of development of the green economy become increasingly important influencing factors. The proposed forecasting approach helps the regional electricity generation firms reduce a large amount of carbon dioxide emissions.

Parameterization modeling for wind drift factor in oil spill drift trajectory simulation based on machine learning

Marine oil spill simulations typically employ the oil particle method to calculate particle trajectories, considering various factors such as wind, current, and turbulence. The wind drift factor (WDF), a random element determining the proportion of wind’s effect on oil particles, is often empirically set as a constant in traditional oil spill models, introducing limitations. This study proposes a support vector regression-based parameterization modeling (SVR-PM) for the WDF. Using extensive buoy data and ocean hydrodynamic reanalysis data, we trained an SVR model to compute the WDF in real-time based on real-time wind speed. The SVR-PM was integrated into an oil spill model to enhance the computation of the wind-induced velocity term. We validated the model using satellite images of two significant oil spills, resulting in an excellent average agreement. The SVR-PM’s advantage lies in enhancing the accuracy of wind-induced velocity term in oil spill simulations and demonstrating strong adaptability and generalizability over time and space. This advancement holds significant implications for maritime departments and emergency disaster response units.

Water Inflow Amount Prediction for Karst Tunnel with Steady Seepage Conditions

Tunnel engineering is an important aspect of developing and utilizing underground spaces. Tunnel water inrush became a common problem that restricts the safe and efficient construction of tunnels. This paper focuses on a karst water-rich tunnel in Chongqing and establishes a seepage field distribution model around the tunnel, analyzing the evolution law of the seepage field. The water balance method and underground runoff modulus method are used to predict tunnel water inflow. The prediction method for tunnel water inflow in water-rich karst areas is combined with long-term on-site tunnel hydrology observations. The distribution of groundwater in front of the tunnel face is drawn using the software, successfully predicting the larger karst area in front of the face. The prediction of water inrush risk level for karst tunnels is carried out using the SVR model. An expression formula for the water head around the tunnel is established by using the conformal mapping relationship, and the distribution status of the seepage field around the tunnel is ultimately determined. The overall prediction accuracy of the underground runoff modulus method is better than that of the water balance method in predicting the water inrush volume of the tunnel. The prediction of the large karst area ahead of the heading is successfully achieved by using the SVR model. This prediction method can provide reference and guidance for the construction of other karst-rich water tunnels in the region.

A steps-ahead tool wear prediction method based on support vector regression and particle filtering

This paper develops a steps-ahead tool wear prediction method based on particle filtering and support vector regression. A degradation phase classification method is presented based on clustering algorithm and support vector machine. The support vector regression models are established to achieve the mapping between degradation features and the flank tool wear values. In online prediction, the measured signals are input into the SVM model to judge the current phase of the tool, and the particle filtering algorithm is used for online steps-ahead prediction of the features. The prediction of tool wear can be obtained by inputting the feature prediction results into the SVR models. The experimental tool wear dataset is introduced as an application example to test the proposed method. The results demonstrate the effectiveness of the proposed method, and the comparison with other tool wear prediction shows the advantage of the proposed method in prediction accuracy.

Emerging pathways to sustainable economic development: An interdisciplinary exploration of resource efficiency, technological innovation, and ecosystem resilience in resource-rich regions

Innovation has always been a key factor in economic growth. But with the increasing pressure on natural resources, it has become even more crucial to understand the relationship between innovation, resources, and economic development. Technology has also played a significant role in helping to achieve sustainability. By utilizing the right strategies and investing in innovative technologies, businesses and governments can promote sustainable economic progress while decreasing their influence on the environment. This work explores the nexus among resource consumption, technological innovations, sustainability, and economic development and how this relationship has evolved from a traditional resource-poor region to a more resource-rich one. A correlation heat map showed the highest association between GDP growth (GDPG) and research and development expenditure (RDE) (−0.47). Support vector Regression with radial basis function (SVR-RBF) was applied to see the impact of contributing parameters that influence GDPG. For GDPG, the model generated R = 0.88 and R2 = 0.78 (RMSE = 1.16). The results of the SVR showed that the strongest coefficient in the SVR model is self-employed, total (SE), followed by electricity production from the nuclear resource (EPNR). This study could have several contributions. Firstly, the study’s focus on sustainable economic development provides insights into how resource-rich regions can balance economic growth with environmental conservation. This information could be valuable to policymakers, businesses, and other stakeholders interested in creating sustainable economic growth strategies. Secondly, the study’s interdisciplinary approach could help bridge the gap between different fields of study, leading to more holistic and effective solutions to environmental and economic challenges. This interdisciplinary approach could also help build stronger partnerships between academia, industry, and government. Finally, the study’s exploration of resource efficiency, technological innovation, and ecosystem resilience could provide useful information for China, a resource-rich country undergoing rapid economic growth. China faces numerous environmental and economic challenges, and the study’s findings could help inform policies and strategies to address these challenges.

Eocene Statistics of the Semi-Deep-Deep Lake-Level Evolution of the Jijuntun Formation in the Fushun Basin, Northeast China.

The fluctuation of lake levels in semi-deep and deep lake environments has long been a central topic in the study of ancient lake evolution. This phenomenon has a significant impact on the enrichment of organic matter and the overall ecosystem. The study of lake-level changes in deep lake environments is hindered by the scarcity of records in continental strata. To address this issue, we conducted a study on the Eocene Jijuntun Formation in Fushun Basin, specifically focusing on the LFD-1 well. Our study involved finely sampling the extremely thick oil shale (about 80 m), which was deposited in the semi-deep to deep lake environment of the Jijuntun Formation. The TOC was predicted by multiple methods, and the lake level study was restored by combining logging INPEFA and Dynamic noise after orbital tuning (DYNOT) techniques. The oil shale of the target layer is type I kerogen, and the source of organic matter is basically the same. The γ ray (GR), resistivity (RT), acoustic (AC), and density (DEN) logging curves are in the normal distribution, indicating that the logging data are better. The accuracy of TOC simulated by improved Δlog R, SVR, and XGBoost models is affected by the number of sample sets. The improved Δlog R model is most affected by the change of sample size, followed by the SVR model, and the XGBoost model is the most stable. In addition, compared with the prediction accuracy of TOC by improved Δlog R, SVR, and XGBoost models, it is shown that the improved Δlog R method has limitations in the prediction of TOC in oil shale. The SVR model is more suitable for the prediction of oil shale resources with small sample size, and the XGBoost model is applicable when the sample size is relatively large. According to the DYNOT analysis of logging INPEFA and TOC, the lake level changes frequently during the deposition of ultra-thick oil shale, and the lake level has experienced five stages of rising-stabilizing-frequent fluctuation-stabilizing- decreasing. The research results provide a theoretical basis for revealing the plane change of stable deep lake lakes and provide a basis for the study of lake levels in faulted lake basins in Paleogene Northeast Asia.

Leaf Spectral Analysis for Detection and Differentiation of Three Major Rice Diseases in the Philippines

Monitoring the plant’s health and early detection of disease are essential to facilitate effective management, decrease disease spread, and minimize yield loss. Spectroscopic techniques in remote sensing offer less laborious methods and high spatiotemporal scale to monitor diseases in crops. Spectral measurements during the development of disease infection may reveal differences among diseases and determine the stage it can be effectively detected. In this study, spectral analysis was performed over the visible and near-infrared (400–850 nm) portions of the spectrum to detect and differentiate three major rice diseases in the Philippines, namely tungro, BLB, and blast disease. Reflectance of infected rice leaves was recorded repeatedly from inoculation to the late stage of each disease. Results show that spectral reflectance is characteristically affected by each disease, resulting in different spectral, signature sensitivity, and first-order derivatives. Red and red-edge wavelength ranges are the most sensitive to the three diseases. Near-infrared wavelengths decreased as tungro and blast diseases progressed. In addition, the spectral reflectance was resampled to common reflectance sensitivity bands of optical sensors and used in the cluster analysis. It showed that BLB and blast can be detected in the early disease stage on the IRRI Standard Evaluation System (SES) scale of 1 and 3, respectively. Alternatively, tungro was detected in its later stage, with an 11–30% height reduction and no distinct yellow to yellow-orange discoloration (5 SES scale). Three regression techniques, Partial Least Square, Random Forest, and Support Vector Regression were performed separately on each disease to develop models predicting its severity. The validation results of the PLSR and SVR models in tungro and blast show accuracy levels that are promising to be used in estimating the severity of the disease in leaves while RFR shows the best results for BLB. Early disease detection and regression models from spectral measurements and analysis for disease severity estimation can help in disease monitoring and proper disease management implementation.

Flood Simulations Using a Sensor Network and Support Vector Machine Model

This study aims to couple the support vector machine (SVM) model with a hydrometeorological wireless sensor network to simulate different types of flood events in a montane basin. The model was tested in the mid-latitude montane basin of Vydra in the Šumava Mountains, Central Europe, featuring complex physiography, high dynamics of hydrometeorological processes, and the occurrence of different types of floods. The basin is equipped with a sensor network operating in headwaters along with the conventional long-term monitoring in the outlet. The model was trained and validated using hydrological observations from 2011 to 2021, and performance was assessed using metrics such as R2, NSE, KGE, and RMSE. The model was run using both hourly and daily timesteps to evaluate the effect of timestep aggregation. Model setup and deployment utilized the KNIME software platform, LibSVM library, and Python packages. Sensitivity analysis was performed to determine the optimal configuration of the SVR model parameters (C, N, and E). Among 125 simulation variants, an optimal parameter configuration was identified that resulted in improved model performance and better fit for peak flows. The sensitivity analysis demonstrated the robustness of the SVR model, as different parameter variations yielded reasonable performances, with NSE values ranging from 0.791 to 0.873 for a complex hydrological year. Simulation results for different flood scenarios showed the reliability of the model in reconstructing different types of floods. The model accurately captured trend fitting, event timing, peaks, and flood volumes without significant errors. Performance was generally higher using a daily timestep, with mean metric values R2 = 0.963 and NSE = 0.880, compared to mean R2 = 0.913 and NSE = 0.820 using an hourly timestep, for all 12 flood scenarios. The very good performance even for complex flood events such as rain-on-snow floods combined with the fast computation makes this a promising approach for applications.

A Surrogate Machine Learning Model for the Design of Single-Atom Catalyst on Carbon and Porphyrin Supports towards Electrochemistry

We apply the machine learning (ML) tool to calculate the Gibbs free energy (ΔG) of reaction intermediates rapidly and accurately as a guide for designing porphyrin- and graphene-supported single-atom catalysts (SACs) toward electrochemical reactions. Based on the 2105 DFT calculation data from the literature, we trained a support vector machine (SVR) algorithm. The hyperparameters were optimized using Bayesian optimization along with 10-fold cross-validation to avoid overfitting. Based on the Shapley Additive exPlanation (SHAP) and permutation methods, the feature importance analysis suggests that the most important parameters are the number of pyridinic nitrogen (Npy), the number of d electrons (θd), and the number of valence electrons of reaction intermediates. Inspired by this feature importance analysis and the Pearson correlation coefficient, we found a linear dependent, simple, and general descriptor (φ) to describe ΔG of reaction intermediates (e.g., ΔGOH* = 0.020φ – 2.190). Using the trained SVR algorithm, ΔGOH*, ΔGO*, ΔGOOH*, ΔGOO*, ΔGH*, ΔGCOOH*, ΔGCO*, and ΔGN2* intermediates are predicted for the oxygen reduction reaction (ORR), the oxygen evolution reaction (OER), the hydrogen evolution reaction (HER), and the CO2 reduction reaction (CO2RR). The SVR model predicts an ORR overpotential of 0.51 V and an HER overpotential of 0.22 V for FeN4-SAC. Moreover, we used the SVR algorithm for high-throughput screening of SACs, suggesting new SACs with low ORR overpotentials. This strategy provides a data-driven catalyst design method that significantly reduces the costs of DFT calculations while providing the means for designing SACs for electrocatalysis and beyond.

Soil salinity estimation in Shule River Basin using support vector regression model

AbstractSoil salinization is a progressive degradation process that spreads globally and leads to a decline in soil fertility. Assessing the scale of salinization is crucial for sustainable agricultural development and saline‐land rehabilitation. In this study, we proposed a support vector regression spatial (SVR‐S) model that utilizes spatial dependence information to predict soil salinity over the irrigation area of the Shule River Basin in northwestern China. To investigate the performance of the SVR‐S model, 50 soil samples were collected in the field. Semivariograms of soil salinity (measured as total salt content and ion concentrations) were constructed to measure spatial dependence. The SVR‐S model was compared with the original SVR model and the geographically weighted regression (GWR) model regarding the salinity prediction ability. The soil salinity in the experimental area demonstrated a strong spatial dependence pattern. The SVR‐S model delivered a better performance than the SVR‐O and GWR. SVR‐S showed a correlation coefficient R of 0.87 and a root mean square error (RMSE) of 1.83%, while the performance of SVR‐O (R = 0.75, RMSE = 3.32%) and GWR (R = 0.73, RMSE = 3.47%) was comparatively poor. Topographic indices integrating spatial information contributed the most to the estimation of salinity in the study area. This study provides a new approach to integrating spatial information for accurate soil salinity mapping.

Python Model Predicts Covid-19 Cases since Omicron in Indonesia

The proposed work uses Support Vector Regression model to predict the new cases, recovered cases, and deaths cases of covid-19 every day during sub-variant omicron spread in Indonesia. We collected data from June 14, 2022, to August 12, 2022 (60 Days). This model was developed in Python 3.6.6 to get the predictive value of the issues mentioned above up to September 21, 2022. The proposed methodology uses a SVR model with the Radial Basis Function as the kernel and a 10% confidence interval for curve fitting. The data collected has been divided into 2 with a size of 40% test data and 60% training data. Mean Squared Error, Root Mean Squared Error, Regression score, and percentage accuracy calculated the model performance parameters. This model has an accuracy above 87% in predicting new cases and recovered patients and 68% in predicting daily death cases. The results show a Gaussian decrease in the number of cases, and it could take another 4 to 6 weeks for it to drop to the minimum level as the origin of the undiscovered omicron sub-variant. RBF (Radial Basis Function) very efficient and has higher accuracy than linear or polynomial regression as kernel of SVR.

Contributions to auditory system conduction velocity: insights with multi-modal neuroimaging and machine learning in children with ASD and XYY syndrome.

The M50 electrophysiological auditory evoked response time can be measured at the superior temporal gyrus with magnetoencephalography (MEG) and its latency is related to the conduction velocity of auditory input passing from ear to auditory cortex. In children with autism spectrum disorder (ASD) and certain genetic disorders such as XYY syndrome, the auditory M50 latency has been observed to be elongated (slowed). The goal of this study is to use neuroimaging (diffusion MR and GABA MRS) measures to predict auditory conduction velocity in typically developing (TD) children and children with autism ASD and XYY syndrome. Non-linear TD support vector regression modeling methods accounted for considerably more M50 latency variance than linear models, likely due to the non-linear dependence on neuroimaging factors such as GABA MRS. While SVR models accounted for ~80% of the M50 latency variance in TD and the genetically homogenous XYY syndrome, a similar approach only accounted for ~20% of the M50 latency variance in ASD, implicating the insufficiency of diffusion MR, GABA MRS, and age factors alone. Biologically based stratification of ASD was performed by assessing the conformance of the ASD population to the TD SVR model and identifying a sub-population of children with unexpectedly long M50 latency. Multimodal integration of neuroimaging data can help build a mechanistic understanding of brain connectivity. The unexplained M50 latency variance in ASD motivates future hypothesis generation and testing of other contributing biological factors.

Bead morphology prediction of coaxial laser cladding on inclined substrate using machine learning

Laser cladding technology has been widely used in the repair and surface strengthening of huge parts. In order to avoid interference between the laser head and the surface to be deposited, the laser cladding process needs to be completed in a tilted position in order to remanufacture some complex parts. The morphology of the cladding layer directly affects the overall quality of the surface after laser cladding. To solve the problem that the quality of the coating cannot be guaranteed under an inclined attitude, SVR, PSO-BPNN, and XGBoost models were applied to build a prediction model between the bead morphology and process parameters. An L36 orthogonal array with four factors and three levels was used to obtain the original data set, and the four-fold cross-validation method was used to train and validate the three models. The predictive performance showed that the XGBoost model had better performance in predicting the morphology of the cladding layer on the inclined substrate compared to the other two algorithms. The R2 values of the width, height, and peak shifting point prediction model using XGBoost were 0.9578, 0.9419, and 0.8765 respectively. Finally, the prediction accuracy of the XGBoost model was evaluated by using other 9 test data sets. The maximum error in width was 3.73 %, and the errors in height and peak shifting point were less than 10 %. The prediction model also performed well in terms of MAE, RMSE, and R2 by using other nine test data sets, further demonstrating the generalization capability of XGBoost. This prediction model can be used to optimize the process path of la-ser cladding with a tilted position and effectively improve the processing efficiency and quality of complex surface laser cladding.

Lithium-Ion Battery Health State Prediction Based on VMD and DBO-SVR

Accurate estimation of the state-of-health (SOH) of lithium-ion batteries is a crucial reference for energy management of battery packs for electric vehicles. It is of great significance in ensuring safe and reliable battery operation while reducing maintenance costs of the battery system. To eliminate the nonlinear effects caused by factors such as capacity regeneration on the SOH sequence of batteries and improve the prediction accuracy and stability of lithium-ion battery SOH, a prediction model based on Variational Modal Decomposition (VMD) and Dung Beetle Optimization -Support Vector Regression (DBO-SVR) is proposed. Firstly, the VMD algorithm is used to decompose the SOH sequence of lithium-ion batteries into a series of stationary mode components. Then, each mode component is treated as a separate subsequence and modeled and predicted directly using SVR. To address the problem of difficult parameter selection for SVR, the DBO algorithm is used to optimize the parameters of the SVR model before training. Finally, the predicted values of each subsequence are added and reconstructed to obtain the final SOH prediction. In order to verify the effectiveness of the proposed method, the VMD-DBO-SVR model was compared with SVR, Empirical Mode Decomposition-Support Vector Regression (EMD-SVR), and VMD-SVR methods for SOH prediction of batteries based on the NASA dataset. Experimental results show that the proposed model has higher prediction accuracy and fitting degree, with prediction errors all within 1% and better robustness.

State of health estimation for lithium-ion battery based on improved support vector regression

The lithium-ion battery has grown to be one of the most popular types of energy storage because of its many great qualities. However, the state of health (SOH) of the lithium-ion battery decreases as the number of cycles increases, resulting in reduced performance or even failure of the device, so an accurate SOH is essential for the safe operation of the device. To solve the issue of existing estimate methods' low estimation accuracy, a SOH estimation approach based on improved support vector regression is proposed. Firstly, the changes in battery characteristics during charging and discharging are analyzed and the health factors characterizing the SOH degradation are extracted. Pearson and Spearman correlation coefficients are used to quantitatively analyze the correlation between SOH and health factors. In addition, by optimizing the kernel function parameters of the SVR model by IALO, the improved ant lion optimization and support vector regression (IALO-SVR) based SOH estimation technique was developed. The IALO-SVR method was validated with the NASA battery dataset, and the experimental results showed that the method can estimate SOH more accurately compared with Back Propagation Neural Network (BP) and ALO-SVR methods, with an estimation error of no more than 1.8% at most.

Ultra-Short-Term Wind Power Prediction Based on LSTM with Loss Shrinkage Adam

With the rapid increase in wind power, its strong randomness has brought great challenges to power system operation. Accurate and timely ultra-short-term wind power prediction is essential for the stable operation of power systems. In this paper, an LsAdam–LSTM model is proposed for ultra-short-term wind power prediction, which is obtained by accelerating the long short-term memory (LSTM) network using an improved Adam optimizer with loss shrinkage (LsAdam). For a specific network topology, training progress heavily depends on the learning rate. To make the training loss of LSTM shrink faster with standard Adam, we use the past training loss-changing information to finely tune the next learning rate. Therefore, we design a gain coefficient according to the loss change to adjust the global learning rate in every epoch. In this way, the loss change in the training process can be incorporated into the learning progress and a closed-loop adaptive learning rate tuning mechanism can be constructed. Drastic changes in network parameters will deteriorate learning progress and even make the model non-converging, so the gain coefficient is designed based on the arctangent function with self-limiting properties. Because the learning rate is iteratively tuned with past loss-changing information, the trained model will have better performance. The test results on a wind turbine show that the LsAdam–LSTM model can obtain higher prediction accuracy with much fewer training epochs compared with Adam–LSTM, and the prediction accuracy has significant improvements compared with BP and SVR models.