Input Variable Selection Research Articles

Forecasting Oil Prices with Non-Linear Dynamic Regression Modeling

The recent energy crisis has renewed interest in forecasting crude oil prices. This paper focuses on identifying the main drivers determining the evolution of crude oil prices and proposes a statistical learning forecasting algorithm based on regression analysis that can be used to generate future oil price scenarios. A combination of a generalized additive model with a linear transfer function with ARIMA noise is used to capture the existence of combinations of non-linear and linear relationships between selected input variables and the crude oil price. The results demonstrate that the physical market balance or fundamental is the most important metric in explaining the evolution of oil prices. The effect of the trading activity and volatility variables are significant under abnormal market conditions. We show that forecast accuracy under the proposed model supersedes benchmark specifications, including the futures prices and analysts’ forecasts. Four oil price scenarios are considered for expository purposes.

ЭФФЕКТИВНОСТЬ АЛГОРИТМА ДЕРЕВА ПРИНЯТИЯ РЕШЕНИЙ В МОДЕЛИ ANFIS НА ПРИМЕРЕ ВОССТАНОВЛЕНИЯ НЕДОСТАЮЩИХ ДАННЫХ

Missing data in water resources studies prevent planning. For this reason, data estimation studies are carried out. In this study, ANFIS (Adaptive Neural Fuzzy Inference System) was used to complete the missing data. At the study area, the Yesilirmak Basin located in the north of Turkey, input variables from seven stations and output variable from one station were determined. In the research, 80% (378 months of data) of 504 months of the flow data between 1969 and 2011 was used in the training phase and 20% (126 months of data) was employed in the testing one. The decision tree was used instead of the trial and error method in the selection of input variables and determining the number of membership functions in ANFIS models. It was concluded that the ANFIS model established with the information obtained from the decision tree is successful compared to the randomly established ANFIS models. Using the decision tree before ANFIS models are created will not only minimize the time spent on the model development, but also prevent the best of the possible models from being overlooked.

Evaluating nano-metal oxide mixed matrix membranes for whey protein separation using hybrid intelligent optimization learning

Ultrafiltration (UF) membranes offer cost-efficient and sustainable techniques for the extraction of whey protein in cheese whey wastewater. This strategy substantially aids in waste minimization and fosters the progression of a circular economy paradigm within the dairy industry. The emerging discipline of machine learning has benefited UF by enhancing the separation performance and deepening the filtration mechanism. In this study, adaptive neuro-fuzzy inference systems (ANFIS) coupled chemometric learning was implemented for the performance evaluation of whey protein recovery in nano metal oxide embedded mixed matrix membranes (MMMs) UF from whey wastewater. The selected input variables are nano metal oxides concentration (NC), pore radius (S), and contact angle (CA). The targeted responses for the parametric dependency analysis are individual whey protein permeability (lysozyme (PLys), lactalbumin (PLA) and bovine serum albumin (PBSA)) and protein rejection (%) (lysozyme (RLys), lactalbumin (RLA) and bovine serum albumin (RBSA)). The Jarque-Bera indicated that the input variables did not deviate from normality, and S and CA have significantly influenced the whey protein rejection and permeability. ANFIS models demonstrate superior predictive capabilities with minimal error for flux under various transmembrane pressure (TMP) in MMMs due to their inherent gradient descent optimization. Among the proteins, a Pearson correlation coefficient approaching unity denotes a significant positive relationship between lysozyme flux (FLys). The ANFIS models exhibited excellent predictive performance in real-time cheese whey effluent flux and rejection data, achieving a mean square error of 0.0003 for experimental flux. This study provides valuable insights for the development of models tailored to small datasets, offering a foundation for predicting and enhancing membrane performance.

Automated Input Variable Selection for Analog Methods Using Genetic Algorithms

AbstractAnalog methods (AMs) have long been used for precipitation prediction and climate studies. However, they rely on manual selections of parameters, such as predictor variables and analogy criteria. Previous work showed the potential of genetic algorithms (GAs) to optimize most of the AM parameters. This research goes one step further and investigates the potential of GAs for automating the selection of the input variables and the analogy criteria (distance metric between two data fields) in AMs. Our study focuses on the prediction of daily precipitation in central Europe, specifically Switzerland, as a representative case. Comparative analysis against established methods demonstrates the superiority of GA‐optimized AMs in terms of predictive accuracy. The selected input variables exhibit strong associations with key meteorological processes that influence the generation of precipitation. Further, we identify a new analogy criterion inspired by the Teweles‐Wobus criterion, which consistently performs better than other Euclidean distances and could be used in classic AMs. In contrast to conventional stepwise selection approaches, GA‐optimized AMs display a preference for a flatter structure characterized by a single level of analogy and an increased number of variables. Overall, our study demonstrates the successful application of GAs in automating input variable selection for AMs, with potential implications for application in diverse locations and data exploration to predict alternative predictands. In a broader context, GAs could be used to perform input variable selection in other data‐driven methods, opening perspectives for a broad range of applications.

A STATISTICAL ANALYSIS OF GEOLOGICAL AND ENGINEERING PREDICTORS OF OILFIELD PERFORMANCE RESPONSE: A CASE STUDY OF OILFIELDS ON THE UK CONTINENTAL SHELF

Oilfield production is controlled by a wide range of geological and engineering parameters, many of which are at least partially interrelated. This paper uses multivariate statistical methods (principal component analysis, regression analysis and analysis of variance) to determine how these parameters are related, and which of them are most significant in controlling and predicting oilfield performance. The analysis is based on a database of publicly available oilfield data from the United Kingdom Continental Shelf (UKCS), from which a series of geological, engineering and fluid‐related control variables from 136 fields were pre‐processed and analyzed. This dataset is a subset of a much wider project database for UKCS oil, gas and condensate fields. For this study, the project database was divided into two datasets: a first dataset with 10 parameters from 136 fields, and a second, more detailed dataset with 27 parameters from 38 fields. Both datasets were analysed using principal component analysis in order to investigate possible correlations between numerically/statistically interrogable predictor variables such as porosity, permeability, number of production wells, gas‐oil ratio and reservoir temperature. A regression analysis was then carried out on the predictor variables in order to obtain a ranking of predictability (i.e. how indicative a predictor is of a particular outcome) and sensitivity (how sensitive an outcome is to slight changes in a predictor) in relation to recovery factor based on R‐squared and regression coefficient values. The results showed that key variables from the principal component analysis included field size, number of production wells, PVT, gross depositional environment and reservoir quality. High‐ranking parameters of predictability and sensitivity from the regression analysis were found to include API, net‐to‐gross, porosity and reservoir depth. These results are consistent with previous studies and suggest that it should be possible to forecast oilfield recovery based on only a few selected input variables. As a preliminary test of forecasting ability of the variable permutations put forward, a best‐subsets multiple regression was carried out using a statistical software package and yielded results which corroborated the main findings.

Analysis of dissolved oxygen influencing factors and concentration prediction using input variable selection technique: A hybrid machine learning approach

Accurate quantification of dissolved oxygen (DO) is critically important for the protection and management of aquatic ecosystems. Successful applications have utilized mechanistic and data-driven models to simulate DO content in aquatic ecosystems. However, mechanistic models present challenges due to their complex and difficult-to-solve conditions, making them less portable. Additionally, data-driven model predictions are hindered by the challenge of numerous input variables, impacting both the running speed and prediction performance of the model. To address these challenges, water quality data and meteorological data of the Tanjiang River were obtained. The maximum information coefficient (MIC) input variable selection technique was employed to identify primary environmental factors influencing DO changes. Furthermore, coupled with support vector regression (SVR), two models (SVR and MIC-SVR) were employed to estimate the DO concentration of the Tanjiang River, and the optimal model was established. The results indicated a shift in the primary pollution factor from ammonia nitrogen to total phosphorus after recent treatment in the Tanjiang River. In comparison with the SVR model, the root mean square error (RMSE) of the MIC-SVR model was reduced by 4.46%, and the Nash-efficiency coefficient (NSE) was improved by 45.85%. In addition, study of kernel function selection revealed that considering as many kernel functions as possible is necessary for improving the performance of the SVR model. Conclusively, the proposed MIC-SVR model serves as an effective tool to analyze the relationship between DO and environmental factors, identifying the primary causes of low DO, and accurately predict the DO concentration in the Tanjiang River (especially in its middle and lower reaches), thus providing a reference for governmental decision-making on water environmental protection and water resource management.

Short term wind speed forecasting using artificial and wavelet neural networks with and without wavelet filtered data based on feature selections technique

Wind speed forecasting plays a crucial role in enhancing the efficiency, reliability, and profitability of renewable energy systems. Accurate wind speed forecasting optimizes energy production and grid integration in renewable energy systems. It assists in maintenance scheduling, decision-making in energy markets, and supports risk management for financial planning. This paper investigates the difficult balance between model complexity and forecasting accuracy in wind speed forecasting using Artificial Neural Networks (ANNs) and Wavelet Neural Networks (WNNs). Employing a time series approach, the study develops models for 1-h-ahead wind speed forecasting, utilizing recent averaged data and exploiting correlations between consecutive wind speeds for improved short-term predictions. Models training involve the Backpropagation algorithm, with careful input variable selection to minimize errors. This study emphasizes that simpler ANN and WNN models can outperform complex ones and explores the effectiveness of wavelet filtering techniques and demonstrates the benefits of using wavelet filtering to enhance forecasting accuracy. The robustness of the developed models is validated through K-fold cross-validation, confirming the efficacy of the proposed models. Evaluation across two datasets demonstrates the superiority of the proposed models over the persistence model, with enhancements of 7–14% in RMSE. Additionally, the efficacy of wavelet transforms in enhancing forecasting accuracy compared to regular neural network predictions is emphasized, underscoring the effectiveness of wavelet filtering offering a strategic approach for optimizing wind power management in electrical grids.

UNFIS: A Novel Neuro-Fuzzy Inference System with Unstructured fuzzy rules

An important constraint of Fuzzy Inference Systems (FIS) is their structured rules. Indeed, each rule should evaluate all input variables in its antecedent part. Thus, the length of all fuzzy rules and the number of input variables are equal. However, in many decision-making problems evaluating some conditions on a limited set of input variables is sufficient to decide properly (unstructured rules). Consequently, the FIS’s performance, generalizability, and interpretability are restricted by the aforementioned limitation. This study provides a neuro-fuzzy inference system that can create each fuzzy rule taking into account various sets of input variables in order to handle this problem. To realize this capability, a new fuzzy selector neuron with an adaptive parameter is proposed that can select input variables in the antecedent part of each fuzzy rule. Moreover, the consequent part of the Takagi–Sugeno–Kang FIS is properly changed to consider only the selected set of input variables. To learn the parameters of the proposed architecture, the Levenberg–Marquardt (LM) method is used for regression problems. For classification applications, a trust-region-based learning method (General quasi-Levenberg–Marquardt (GqLM)) is proposed to minimize cross-entropy in multiclass problems. The performance of the proposed method is compared with some related previous approaches in some real-world classification and regression problems. Based on these comparisons, the proposed method has better or very close performance with a parsimonious structure consisting of unstructured fuzzy rules.

Monthly Streamflow Prediction of the Source Region of the Yellow River Based on Long Short-Term Memory Considering Different Lagged Months

Accurate and reliable monthly streamflow prediction plays a crucial role in the scientific allocation and efficient utilization of water resources. In this paper, we proposed a prediction framework that integrates the input variable selection method and Long Short-Term Memory (LSTM). The input selection methods, including autocorrelation function (ACF), partial autocorrelation function (PACF), and time lag cross-correlation (TLCC), were used to analyze the lagged time between variables. Then, the performance of the LSTM model was compared with three other traditional methods. The framework was used to predict monthly streamflow at the Jimai, Maqu, and Tangnaihai stations in the source area of the Yellow River. The results indicated that grid search and cross-validation can improve the efficiency of determining model parameters. The models incorporating ACF, PACF, and TLCC with lagged time are evidently superior to the models using the current variable as the model inputs. Furthermore, the LSTM model, which considers the lagged time, demonstrated better performance in predicting monthly streamflow. The coefficient of determination (R2) improved by an average of 17.46%, 33.94%, and 15.29% for each station, respectively. The integrated framework shows promise in enhancing the accuracy of monthly streamflow prediction, thereby aiding in strategic decision-making for water resources management.

Machine Learning and Statistical Test–Based Culvert Condition Impact Factor Analysis

For managers of road infrastructure, culvert deterioration is a major concern since culvert failures can cause serious risks to the traveling public. The efficiency of the cost- and labor-intensive culvert inspection and maintenance process can be improved by properly identifying the key impact factors on culvert condition deterioration. Although the use of machine learning (ML) techniques to predict culvert conditions has been proven to be a promising tool for enhancing culvert management and enabling proactive scheduling of maintenance tasks, the information provided by the developed ML models has been given little attention for further use and analysis. By utilizing the predictor importance results of an evaluated decision tree (DT) culvert condition prediction model and the Mann–Whitney U test, this study provided insights to the identification of the key variables influencing culvert deterioration. According to the findings, five impact factors, including culvert span, pH, age, rise, and cover height, often have significant impact on the condition ratings of culverts made of various materials. In addition, such a statistical test-assisted factor identification process offered a way of identifying and enhancing the input variable selection for predictive ML model development.

Probabilistic-Mechanistic Evaluation of a Hybrid Drying System Using l1 Norm Regularisation: Simultaneous Ohmic Heating and Convection Drying

Ohmic-assisted drying (OAD) is a novel drying system that combines ohmic heating and convection drying simultaneously. The present study is aimed at evaluating the mechanism of OAD system behaviours against the combined impact of operational and model uncertainties. Moreover, the dynamic (time-dependent), as well as static (end-of-drying) spatial homogeneity of the model predictions, was quantitatively described for the first time in the literature using Buzas and Gibson’s evenness value (an α-diversity index). The Monte Carlo simulation approach was used to propagate the uncertainty of randomly selected input variables using the Halton sequence sampling method. Mechanistic models include input uncertainties that lead to deviations in model estimations. Both time-dependent and independent global sensitivity of moisture, sample temperature, sample’s internal pressure, their spatial homogeneity, and drying time were assessed using Lasso regression (a variable selection method that penalises the coefficients using l1 norm). The stochastic results of the mechanistic investigation showed that the effects of the input variables are almost identical both as static and time-dependent variables. Lasso regression results indicated that operational and model uncertainties cause varying changes in the magnitude and direction of the stochastic model predictions throughout the process. By contrast, the homogeneity properties of the dry product are not caused by these variations and heterogeneous distribution of the electric field. Additionally, electrical conductivity, oven temperature, applied voltage, and initial moisture were found to be the variables which have the most significant effect on all the variables which were examined in terms of operational and model uncertainties. Practical Applications. The present study investigates the stochastic behaviour of the OAD system through the mechanistic model and using a probabilistic modelling approach. To the best of our knowledge, in addition to being the first study to probabilistically evaluate the OAD system, we are introducing Buzas and Gibson’s evenness value for the first time in the food science/technology literature. This measure serves as a numerical indicator of the spatial distribution homogeneity of the physical properties of the sample. The study’s findings and proposed methodologies will have further applications not only for researchers but also for manufacturers, particularly for those involved in the design and analysis of new drying and food systems in general. Moreover, the presented methods and novel homogeneity measures are generic tools; they can be easily adapted to other process improvement practices involving input/output uncertainty/variability.

A hybrid soft sensor for key product yield of FCC unit based on deep learning framework driven by data and process mechanism

The model for online prediction of product yield embedded in the optimization and control framework has the potential to improve the profits of fluid catalytic cracking (FCC) units. In this work, a hybrid soft sensor model is developed to predict key product yield in commercial FCC units. The deep learning framework based on semi-supervised learning is driven by data and process mechanism to construct the hybrid model with a double-layer structure. Moreover, process simulation strategy is utilized for key input variable selection of the hybrid model to ensure model reliability. The hybrid model is compared to a pure data-driven model in terms of accuracy and trend consistency of model predictions. The results demonstrate that the hybrid model exhibits a superior prediction performance. This work suggests that the hybrid model can drive innovation and industrial practice in the optimization and control of FCC units.

A deep learning modeling framework with uncertainty quantification for inflow-outflow predictions for cascade reservoirs

Accurate prediction of reservoir inflows and outflows and their uncertainties is essential for managing water resources and establishing early-warning systems. However, this can be a formidable challenge due to numerous uncertainties, particularly in cascade reservoir systems. To seamlessly quantify aleatoric (data-caused) and epistemic (model network–caused) uncertainties simultaneously in a single framework, we estimated the posterior distribution of parameters in a Bayesian neural network while measuring prediction variance to reflect the noise in data. By randomly discarding certain units within the network, the estimated posterior distribution can be combined with a new loss function. This sophisticated deep learning framework for a long short-term memory network has also included advanced supporting approaches, such as input variable selection, data transformation, and hyperparameter optimization. The model trained in this study was applied to two cascade reservoir systems with four reservoirs in Vietnam, providing uncertainty estimates from two distinct sources. Comparing three global reservoir operation schemes, we found that the proposed model achieved superior performance in all cases, and uncertainty in the forecasts was due primarily to data noise, rather than uncertainty in the model itself. Preprocessing data using a wavelet transform can reduce noise that the model cannot classify on its own, resulting in improved performance, particular with longer lead times. The satisfactory performance of the prediction results confirms that the framework can effectively assess the uncertainty of hydrologic predictions using deep learning.

Predicting blast furnace permeability index: a deep learning approach with limited time-series data

The blast furnace permeability index is one of the crucial technical indicators in the ironmaking process of a blast furnace. Given that the conventional models are not entirely suitable for accommodating the intricate characteristics of blast furnace production, this paper explores a comprehensive approach that involves data mining, the sparrow search algorithm (SSA), convolutional neural networks (CNNs), and gated recurrent unit networks (GRUs) for predicting the blast furnace permeability index. Initially, to address the multi-noise nature of blast furnaces, outliers are eliminated, and a Kalman filter is devised for denoising purposes. Subsequently, in consideration of the nonlinear and substantial time-delay features of blast furnaces, the maximal information coefficient (MIC) method is employed for time-delay alignment, followed by the selection of model input variables based on process analysis and relevance. Subsequent to this, the SSA-CNN-GRU model is established. Within the modeling process, a one-dimensional convolutional neural network is utilized to extract distinct process variable features, thus further resolving the interdependence among blast furnace data. Ultimately, the effectiveness, accuracy, and advancement of the proposed method are validated using real production data.

Reference Equations for Within-Breath Respiratory Oscillometry in White Adults.

Within-breath analysis of oscillometry parameters is a growing research area since it increases sensitivity and specificity to respiratory pathologies and conditions. However, reference equations for these parameters in White adults are lacking and devices using multiple sinusoids or pseudorandom forcing stimuli have been underrepresented in previous studies deriving reference equations. The current study aimed to establish reference ranges for oscillometry parameters, including also the within-breath ones in White adults using multi-sinusoidal oscillations. White adults with normal spirometry, BMI ≤30 kg/m2, without a smoking history, respiratory symptoms, pulmonary or cardiac disease, neurological or neuromuscular disorders, and respiratory tract infections in the previous 4 weeks were eligible for the study. Study subjects underwent oscillometry (multifrequency waveform at 5-11-19 Hz, Resmon PRO FULL, RESTECH Srl, Italy) in 5 centers in Europe and the USA according to international standards. The within-breath and total resistance (R) and reactance (X), the resonance frequency, the area under the X curve, the frequency dependence of R (R5-19), and within-breath changes of X (ΔX) were submitted to lambda-mu-sigma models for deriving reference equations. For each output parameter, an AIC-based stepwise input variable selection procedure was applied. A total of 144 subjects (age 20.8-86.3 years; height 146-193 cm; BMI 17.42-29.98 kg/m2; 56% females) were included. We derived reference equations for 29 oscillatory parameters. Predicted values for inspiratory and expiratory parameters were similar, while differences were observed for their limits of normality. We derived reference equations with narrow confidence intervals for within-breath and whole-breath oscillatory parameters for White adults.

Enhancing Predictive Accuracy of Insurance Stock Market in Jordan using Hyprid GFS.Thrift Model: A Genetic Fuzzy System-based Fintech Approach

This study focuses on improving the predicting accuracy of the daily ASE's weighted price index of the insurance sector (ICI) using a nonlinear spectral model called maximum overlapping discrete wavelet transform (MODWT) with five mathematical functions, namely, Haar, Daubechies (d4), least square (la8), best localization (bl14), and Coiflet (c6). Using a nonlinear spectral model called maximum overlapping discrete wavelet transform (MODWT) with five mathematical functions—Haar, Daubechies (d4), least square (la8), best localization (bl14), and Coiflet (c6)—this study aims to increase the daily ASE's weighted price index of the insurance sector's (ICI) prediction accuracy. The model utilizes a genetic fuzzy system based on Thrift's methodology (GFS.Thrift). The Amman Stock Exchange (ASE) supplied a dataset with 4,478 observations for the purpose of the study. The dataset represented daily data from January 2, 2006, to March 24, 2024. The adaptive GFS.THRIFT model was trained with 90% of the dataset, while the remaining 10% was used to test its prediction performance. Multiple egressions and multicollinearity tests were used to select input variables such as standardized foreign direct investment (FDI), standardized value traded (VT) and consumer price index (CPI). Insights from this study indicate that all input variables are positively related to the output variable. Secondly, the proposed model (MODWT-Haar-GFS. Thrift) significantly outperforms other existing models including the GFS. Thrift model.

Deep Learning-Based Simulation of Surface Suspended Sediment Concentration in the Yangtze Estuary during Typhoon In-Fa

Effectively simulating the variation in suspended sediment concentration (SSC) in estuaries during typhoons is significant for the water quality and ecological conditions of estuarine shoal wetlands and their adjacent coastal waters. During typhoons, SSC undergoes large variations due to the significant changes in meteorological and hydrological factors such as waves, wind speed, and precipitation, which increases the difficulty in simulating SSC. Therefore, in this study, we use an optimized Principal Component Analysis Long Short-Term Memory (PCA-LSTM) framework with an attention mechanism to simulate the SSC in the Yangtze Estuary during Typhoon In-Fa. First, we integrate data from different sources into a multi-source dataset. Second, we use the PCA to reduce the dimensionality of the multi-source data and eliminate redundant variables in the feature data. Third, we introduce an attention mechanism to optimize the long and short-term memory (LSTM) model. Finally, we use the differential evolution (DE) algorithm for hyperparameter selection and merge the feature data with the SSC data as the input of the optimized LSTM network to simulate SSC. The results showed that SSC’s fitting coefficients (R2) at four hydrological stations improved by 7.5%, 6.1%, 7.4%, and 7.8%, respectively, using the attention-based PCA-LSTM compared to the PCA-LSTM. Moreover, compared to the traditional LSTM model, the R2 was improved by 33.8%, 30.5%, 32.0%, and 28.6%, respectively, using the attention-based PCA-LSTM framework. The study indicates that the selection of input variables can affect the model results. Introducing an attention mechanism can effectively optimize the PCA-LSTM framework and improve the simulation accuracy, which helps simulate the non-linear process of SSC variation occurring during Typhoon In-Fa.

Application of the bayesian model averaging algorithm in evaluating and selecting optimal salinity prediction models

Salinity intrusion poses significant challenges to coastal regions worldwide. Reliable salinity prediction models can provide valuable information to mitigate the impact and influence of salinity intrusion. However, their accuracy relies mainly on the selected input variables and used optimal models. This study employs the BayesianModel Averaging (BMA) algorithm to evaluate input variable importance and select the most reliable salinityprediction model. Based on an analysis of observed salinity data and climate data extracted from Landsat 8 OLI in the Google Earth Engine platform, the BMA algorithm identifies the significance of critical variables andoptimal salinity prediction models. Various statistical metrics, including R-squared, Root Mean Square Error (RMSE), and Mean Absolute Error (MAE) from the Random Forest method, were used to verify the performance of these optimal salinity prediction models. These obtained results offer foundational knowledge and valuable insights for future studies in determining appropriate input variables and selecting the best optimalsalinity prediction model.

Incorporating novel input variable selection method for in the different water basins of Thailand

Selecting appropriate input variables for developing a rainfall prediction model is significantly difficult. The present study proposed an innovative framework for input variable selection (IVS) model bootstrapped long short-term recurrent neural network (BTSP-LSTM-RNN) to identify relevant variables for monthly rainfall forecasting. Monthly meteorological and large-scale climatic variables (LCVs) from 1993 to 2022 at two selected river basins in the northern region of Thailand were used for model development. The proposed BTSP-LSTM-RNN model results were compared with the support vector regression with recursive feature elimination (SVR-RFE) and Gradient boosting (GB) by statistical metrics such as coefficient of determination (R2), mean absolute error (MAE), relative root mean squared error (RRMSE), Pearson’s correlation coefficient (r) and mean absolute percentage error (MAPE). BTSP-LSTM-RNN demonstrated exceptional performance, boasting a higher R2 (0.84), MAE (92.28), RRMSE (10.36) in the Wang basin, and R2 (0.83), MAE (242.60), RRMSE (9.93) in the Nan basin. BTSP-LSTM-RNN also achieved the lowest MAPE of 29.82%. Based on this IVS model results, two input variable combinations (IVCs) were designed. IVC-1 is based on BTSP-LSTM-RNN selection, and IVC-2 is an original set of variables. LSTM-RNN, multi-layer perceptron artificial neural network (MLP-ANN), and ensemble model with bootstrapping on random forest (RF) were employed for monthly prediction. When BTSP-LSTM-RNN's selected input variables from IVC-1 are utilized, the BSTP-RF model demonstrates robust performance. It achieves a high R2 (0.82), a low RRMSE (10.13%) suggests accurate predictions, and the r of 0.91 further supports the model's strong linear relationship with observed rainfall data. Based on prediction model results, the BTSP-LSTM-RNN (IVS) model plays a pivotal role in the selection of input variables for rainfall forecasting and its impact on the performance of prediction models (BSTP-RF, MLP-ANN, and LSTM-RNN). These results consistently underscored the pivotal role of the BTSP-LSTM-RNN IVS model in enhancing the precision and reliability of rainfall predictions.

Input-parameter optimization using a SVR based ensemble model to predict landslide displacements in a reservoir area – A comparative study

For reservoir landslides with limited in situ monitoring data, improving the accuracy of an ensemble prediction model through historical displacement-related factor deconstruction and model optimization is challenging. Unfortunately, even though various support vector regression (SVR) based ensemble prediction models have been proposed for this purpose, the optimal combination of new algorithms and their degree of improved prediction accuracy is still unclear. Based on four evaluation indicators, this paper presents a comparative study of a typical landslide displacement prediction model containing four important nodes (landslide displacement decomposition, inducing factor frequency-component extraction, input parameter selection, and optimization algorithm selection) vital to prediction accuracy. The empirical mode decomposition (EMD) series model was adopted to decompose the landslide displacement and extract the frequency component of the factors that affect the landslide movement. Four trajectory similarity judgment models were used to select input variables for the SVR-based prediction model. Nine swarm intelligence (SI) algorithms were used to help optimize the landslide prediction model. A case study shows that the SVR-based ensemble landslide prediction model works well in predicting the displacement of a slow-moving landslide. The combination of EEMD-CEEMDAN-LCSS-PSO yields the best prediction performance, with MAPE, RMSE, MAE, and R2 values of 0.004, 2.38, 8.64, and 0.96, respectively. The findings of this study can impact further studies of landslide ensemble prediction model optimization with insights into the algorithm selection, parameter setting, evaluation measures, and experimental settings.