Predictive Ability Of Model Research Articles

M&A PREDICTIONS: RECONSIDERING THEIR VALUE, END-USERS, AND METHODOLOGIES

The article explores market participants who may benefit from M&A predictions and how their goals may impact the requirements for M&A predictions. These participants (also called end-users of M&A predictions) are company shareholders considering selling their business, shareholders and company management considering acquiring one or a few other companies, shareholders and company management competing with potential M&A targets or buyers, and advisory firms providing investment banking services in the industries where M&A deals occur. Analyzing their goals while applying M&A predictions, the article concludes that the requirements for M&A predictions can be changed depending on these goals. These end-users may benefit from M&A predictions even if the deals they predict won’t happen. These end-users have the potential to significantly influence the outcome of the M&A events they are predicting. The M&A prediction quality criterion imposed by earlier research - the M&A prediction is correct only when a predicted M&A deal happens - can be relaxed depending on the end-users of M&A predictions and their goals. An M&A prediction will be more valuable for end-users if it includes information on both potential targets and potential buyers. M&A prediction may have a more significant value for end-users if it allows for predicting multiple counterparties for a potential party to an M&A deal. The article analyses the existing theoretical basis behind the M&A predictions and concludes that these theories are insufficient to cover all possible reasons behind the deals from the buyers’ and sellers’ perspectives – additional reasons exist that trigger M&A deals. Also, the existing theories are not always proven by the existing research, showing that their correctness may depend on the context. The article explores the current stance of M&A prediction methodologies, such as: binary state prediction models based on a linear combination of independent variables, starting from the earlier works focused on prediction variables for M&A targets to later works dedicated to adding new company-specific prediction variables of the targets and reflecting the context; alternative computational techniques to predict M&A targets, like non-parametric computational techniques, including machine learning; methodologies to predict M&A buyers; methodologies to predict pairs of buyers and targets, researching the relatedness between them. The article concludes that the M&A prediction methodology shall consider and reflect additional motivations for the M&A deal for targets and buyers and shall always include the context. Predicting only targets seems like a one-sided approach. On the contrary, predicting both parties of the deal seems like a promising prediction methodology. Non-parametric computational techniques based on a broader range of prediction variables, reflecting the motivations of the M&A deal’s parties and the context, look like a promising basic prediction methodology that should be further developed. Testing new M&A prediction methodologies within a specific sector for a longer time looks promising for increasing the robustness of the model's prediction ability. Finally, out-of-sample tests done over a longer time are necessary to check the models’ prediction ability.

Prediction Model of Ocular Metastases in Gastric Adenocarcinoma: Machine Learning-Based Development and Interpretation Study.

Background: Although gastric adenocarcinoma (GA) related ocular metastasis (OM) is rare, its occurrence indicates a more severe disease. We aimed to utilize machine learning (ML) to analyze the risk factors of GA-related OM and predict its risks. Methods: This is a retrospective cohort study. The clinical data of 3532 GA patients were collected and randomly classified into training and validation sets in a ratio of 7:3. Those with or without OM were classified into OM and non-OM (NOM) groups. Univariate and multivariate logistic regression analyses and least absolute shrinkage and selection operator were conducted. We integrated the variables identified through feature importance ranking and further refined the selection process using forward sequential feature selection based on random forest (RF) algorithm before incorporating them into the ML model. We applied six ML algorithms to construct the predictive GA model. The area under the receiver operating characteristic (ROC) curve indicated the model's predictive ability. Also, we established a network risk calculator based on the best performance model. We used Shapley additive interpretation (SHAP) to identify risk factors and to confirm the interpretability of the black box model. We have de-identified all patient details. Results: The ML model, consisting of 13 variables, achieved an optimal predictive performance using the gradient boosting machine (GBM) model, with an impressive area under the curve (AUC) of 0.997 in the test set. Utilizing the SHAP method, we identified crucial factors for OM in GA patients, including LDL, CA724, CEA, AFP, CA125, Hb, CA153, and Ca2+. Additionally, we validated the model's reliability through an analysis of two patient cases and developed a functional online web prediction calculator based on the GBM model. Conclusion: We used the ML method to establish a risk prediction model for GA-related OM and showed that GBM performed best among the six ML models. The model may identify patients with GA-related OM to provide early and timely treatment.

LoSNet: A Tailored Deep Neural Network Framework for Precise Length of Stay Prediction in Disease-Specific Hospitalization

During the COVID-19 pandemic, the healthcare sector faced unprecedented challenges in effectively managing hospital resources. A crucial aspect of resource planning and allocation is the ability to predict the expected length of a patient’s stay. Detecting whether a patient requires extended hospitalization or a shorter stay becomes vital for efficient hospital resource allocation and utilization. This paper aims to build a deep learning-based analytical model named LoSNet that predicts the length of stay for each patient at the time of admission to the Hospital. The early prediction of the length of stay requirement would aid healthcare professionals in optimizing the utility of hospital beds and other resources. In this direction, this paper compares the prediction ability of various machine-learning models including Random Forest, Decision Tree, Logistic Regression, and Naïve Bayes with a customized deep neural network model. The dataset used for this analysis includes ground truth on 3,18,438 patients’ length of stay categorized into eleven classes such as 0-10 days being one class, 11-20 days being another class, and so on to more than 100 days. The methodology employed in this study involves data collection, data transformation, and training LoSNet, a deep neural network with no attention mechanisms. The results indicate impressive performance over other models and a random classifier, with a cross-entropy loss of 1.531 and an accuracy of 0.408 in predicting hospitalization durations in an 11-class classification setup, highlighting the framework’s effectiveness in healthcare management.

Optimized Proportioning Model and Efficacy Analysis of High Performance Concrete in Bridge Construction

Abstract As an important part of the raw materials for building construction, high performance concrete is well known for its many properties. This paper analyzes the prerequisites for high-performance concrete proportioning from the technical point of view and processes involved in designing high-performance concrete proportioning. A particle swarm algorithm combined with a neural network is used to establish the relationship model between concrete strength and proportion, and the constructed PSO-BP model is used for the prediction of concrete mixing ratio strength to get the best material proportion. The prediction ability of the BP neural network model, GP model, and PSO-BP model is compared. In order to test the applicability and reliability of the high-performance concrete formulated by the PSO-BP model in members, static load tests were carried out in this paper on a 20m test single girder to test the strain and deflection of concrete in the control section. The results show that the relative error of the high-performance concrete strength predictions predicted by the PSO-BP model is within 2%. It shows that the optimum high-performance concrete can be formulated using the PSO-BP model. The test beams can meet design and practical engineering requirements in terms of strength and stiffness and can be applied to real bridges. Based on the research and analysis, suggestions have been made for the application of high-performance concrete in bridge construction from four aspects.

Predicting leaf nitrogen content in wolfberry trees by hyperspectral transformation and machine learning for precision agriculture.

Leaf nitrogen content (LNC) is an important indicator for scientific diagnosis of the nutrition status of crops. It plays an important role in the growth, yield and quality of wolfberry. This study aimed to develop new spectral indices (NSIs) and constructed machine learning regression (MLR) models for predicting wolfberry tree LNC. By utilizing four smoothing methods and five mathematic transformation methods, we obtained the original spectral dataset and five spectral transformation datasets for quantitative analysis and model establishment. Subsequently, published vegetation indices (PVIs) were acquired, sensitive wavelengths (SWs) were screened and NSIs were calculated based on SWs. Then MLR models were constructed by combining NSIs from six spectral datasets with three machine learning algorithms. Finally, a comparison was made among the MLR models. The study indicated that the application of mathematical transformation highlighted the differences in spectra, the square root, first-derivative and second-derivative transformation improved the prediction accuracy of MLR models constructed based on NSIs (MLR-NSIs models). However, these transformations had little impact on improving the prediction ability of MLR models constructed based on PVIs (MLR-PVIs models). Additionally, The optimal model for predicting the LNC of wolfberry tree was obtained by using the Random Forest (RF) algorithm combined with NSIs developed by first-derivative transformation spectra. The determination coefficient of validation (Rv2) and ratio of percentage deviation (RPD) was 0.71 and 1.90, respectively. In conclusion, this study has demonstrated that the combination of hyperspectral transformation and machine learning is useful for improving the accuracy of LNC estimation in wolfberry trees.

Construction of 11 metabolic-related lncRNAs to predict the prognosis in lung adenocarcinoma

ObjectiveTo explore the metabolism-related lncRNAs in the tumorigenesis of lung adenocarcinoma.MethodsThe transcriptome data and clinical information about lung adenocarcinoma patients were acquired in TCGA (The Cancer Genome Atlas). Metabolism-related genes were from the GSEA (Gene Set Enrichment Analysis) database. Through differential expression analysis and Pearson correlation analysis, lncRNAs about lung adenocarcinoma metabolism were identified. The samples were separated into the training and validation sets in the proportion of 2:1. The prognostic lncRNAs were determined by univariate Cox regression analysis and LASSO (Least absolute shrinkage and selection operator) regression. A risk model was built using Multivariate Cox regression analysis, evaluated by the internal validation data. The model prediction ability was assessed by subgroup analysis. The Nomogram was constructed by combining clinical indicators with independent prognostic significance and risk scores. C-index, calibration curve, DCA (Decision Curve Analysis) clinical decision and ROC (Receiver Operating Characteristic Curve) curves were obtained to assess the prediction ability of the model. Based on the CIBERSORT analysis, the correlation between lncRNAs and tumor infiltrating lymphocytes was obtained.ResultsFrom 497 lung adenocarcinoma and 54 paracancerous samples, 233 metabolic-related and 11 prognostic-related lncRNAs were further screened. According to the findings of the survival study, the low-risk group had a greater OS (Overall survival) than the high-risk group. ROC analysis indicated AUC (Area Under Curve) value was 0.726. Then, a nomogram with T, N stage and risk ratings was developed according to COX regression analysis. The C-index was 0.743, and the AUC values of 3- and 5-year survival were 0.741 and 0.775, respectively. The above results suggested the nomogram had a good prediction ability. The results based on the CIBERSORT algorithm demonstrated the lncRNAs used to construct the model had a strong correlation with the polarization of immune cells.ConclusionsThe study identified 11 metabolic-related lncRNAs for lung adenocarcinoma prognosis, on which basis a prognostic risk scoring model was created. This model may have a good predictive potential for lung adenocarcinoma.

Two-stage imbalanced learning-based quality prediction method for wheel hub assembly

Wheel hub assembly plays a pivotal role in automobile manufacturing, and its quality directly affects the overall reliability of the automobile transmission system. Accurate prediction of wheel bearing assembly quality can save costs and increase productivity. However, the imbalanced data in wheel hub assembly production led to poor model prediction ability, which poses challenges to existing quality prediction methods. To this end, a two-stage imbalanced learning-based quality prediction method is proposed for wheel hub assembly. Concretely, an improved extremely randomized tree is developed for key process variables adaptive selection in the wheel hub assembly production. Then, a Differential Evolution-based Cost-sensitive AdaBoost (DE-CostAda) method is proposed for quality prediction under imbalanced data scenarios. Finally, a real-world experiment has demonstrated the superiority of the proposed method compared with several machine learning and peer imbalance learning methods. The experimental results also reveal some key process variables related to the final product quality, which provides guidance for process optimization of wheel hub assembly.

Genotype by environment model predictive ability in Miscanthus

AbstractMiscanthus is a genus of perennial grasses native to East Asia that shows promise as a biofuel energy source. Breeding efforts for increasing biofuel capability in this genus have focused on two species, namely M. sinensis (Msi) and M. sacchariflorus (Msa). For these efforts to succeed, it is critical that both Msi and Msa, as well as their interspecific crosses, can be grown at a wide range of latitudes. Therefore, the purpose of this study was to investigate how well existing data from Msi and Msa trials grown at locations throughout the northern hemisphere can train state‐of‐the‐art genomic selection (GS) models to predict genomic estimated breeding values (GEBVs) of dry yield for untested Msi and Msa accessions in untested environments. We found that accounting for genotype by environment interaction in the GS model did not notably improve predictive ability. Additionally, we observed that locations at lower latitudes showed higher predictive ability relative to locations at higher latitudes. These results suggest that it is crucial to increase the number of trial locations at higher latitude locations to investigate the source of this correlation. This will make it possible to train GS models using data from environments that are similar to growing conditions at the locations targeted by Msi and Msa breeders. Such an increase of trial locations in target environments could pave the way toward advancing breeding efforts for overwintering ability in Msi and Msa, and ultimately support the potential of Miscanthus as a biofuel crop.

Prediction ability of discrete element model of loose granular media subjected to complex loadings

The discrete numerical method is recognized today as a powerful method to describe the constitutive behaviour of granular media at both the scale of particles and the scale of the representative elementary volume. This method is also more and more used to address boundary value problems in particular when large deformations may occur or when fluid-solid interactions are involved as, for instance, for internal soil erosion problems. Such discrete numerical models are often calibrated and validated with respect to direct shear tests or drained triaxial compressions. However, even if the models provide satisfying results for these loading paths compared to experimental measurements, their validation with respect to more complex loading paths as the isochoric compression or the path at constant stress deviator still present difficulties, in particular for initially loose granular assemblies. This article investigates the importance of the description of the anisotropy of the initial fabric and of the inter-particle friction law in the simulated responses of loose granular assembly to different kinds of loading paths. It shows how the combination of both can modify importantly the simulated responses to some kinds of loading paths. This investigation is carried out for a numerical discrete model made of spheres by comparison with experimental results on sand. The real shape of particles is considered by introducing rolling resistance at contacts into the model, in order to keep a relatively light model and the possibility to tackle further boundary value problems with a limited computational cost.

Ultrasound assisted aqueous two-phase extraction of polysaccharides from corn stigma: Process optimization, structure characterization, and immunomodulatory activity

The two models of ultrasound assisted aqueous two-phase extraction (UATPE) of polysaccharides from corn stigma were established by response surface methodology (RSM) and artificial neural network (ANN), and compared the prediction ability of RSM and ANN models. Subsequently, the extraction process of corn stigma polysaccharides (CSPs) was optimized by the combination of genetic algorithm (GA) and ANN. The results show that two models could precisely predict the CSPs yield, whereas the prediction ability of ANN was better than RSM. The optimum combination of process parameters was as follows: the ultrasound power of 466 W, extraction temperature of 54 °C, liquid-to-solid ratio of 47 mL/g, and extraction time of 31 min. The highest CSPs yield was 9.89% ± 0.14% under the above parameters, and then the crude CSPs were further purified by DEAE-52 and Sephadex G − 100 chromatography to obtain a homogenous fraction (CSPs-2-SG with a molecular weight of 2.86 × 106 Da) that consisted of Man, GlcA, GalA, Glc, Gal, and Xyl with a molar ratio of 9.14:16.35:11.27:91.08:6.49:4.71. Scanning electron microscopy (SEM) and atomic force microscopy (AFM) demonstrated that CSPs-2-SG showed irregular structures including uneven fragments, some holes, and spherical aggregates on the surface. Finally, the immunomodulatory activity of CSPs-2-SG was evaluated by RAW264.7 cells. CSPs-2-SG could improve viability, phagocytic rate, and NO content of RAW264.7 cells. Moreover, CSPs-2-SG could increase the levels of IL − 1β, IL-6, and TNF-α in RAW264.7 cells. The findings provide important references for utilizing corn stigma as a potential natural immune regulator.

Quantitative analysis of pyrazinamide polymorphs in ternary mixtures by ATR-FTIR and Raman spectroscopy with multivariate calibration

The polymorphism of drugs exists widely in solid chemical drugs. It will affect the physical and chemical properties of drugs, as well as bioavailability. So it is very necessary to establish an quantitative method to improve the quality control level of polymorphic drugs. Attenuated total reflectance-Fourier transform infrared (ATR-FTIR) and Raman spectra have been included in many countries’ pharmacopoeia as the drug polymorph analytical technique, and they have many unique advantages. However, for multiple mixed systems, due to the complexity of optical signals, it is difficult to obtain an ideal content prediction model by classical linear regression, so the application of chemometric methods shows advantages. Pyrazinamide is a typical polymorphism drug, three polymorphic forms (α, δ, γ) were obtained. The model prediction ability of two kinds of spectroscopy combined with three kinds of stoichiometric methods was investigated by orthogonal experiment. On this basis, the influence of different combinations of five data preprocessing methods on improving modeling quality was investigated. In this research, Raman spectra combined with partial least squares (PLS), multiplicative scatter correction (MSC), denoise, median and first derivative at the whole spectral range resulted in a better calibration model. It had a RMSEP of 5.3%, 21.6%, and 20.8% for polymorphs α, δ, and γ, respectively. Several methods were used for preprocessing the spectral data could remove unimportant baseline (offset) interference from samples or to correct scattering effects and emphasize spectral the interesting signals. PLS can derive a few components from the independent variable system. Therefore, it may be an effective method to establish a quantitative model for a multi-polymorphism component mixed system.

The Influence of Demographic Factors, Farmer Knowledge, and Motivational Factors on the Adoption of Agricultural Technology Innovation: A Case Study on Dairy Farmers in South Bangka

This study looks into the complex factors that affect dairy producers in South Bangka's use of agricultural technology. Structural Equation Modeling-Partial Least Squares (SEM-PLS) was used to examine the correlations between farmers' knowledge, motivational factors, technology adoption, and demographic characteristics using a sample of 112 farmers. The results highlight the importance of education and adaptability to change by showing a strong positive correlation between demographic characteristics (age, education) and technological adoption. Farmers' knowledge was found to be a crucial element; a substantial positive association suggests that adoption is positively influenced by a deeper understanding of agricultural technologies. Economic incentives and perceived benefits were found to have significant positive connections with technology adoption, indicating the role that perceived benefits play in encouraging the acceptance of new ideas. The model's significant explanatory and predictive ability was validated by the SEM-PLS study, offering a thorough framework for comprehending and improving South Bangka's adoption of technology. The findings highlight the need of a comprehensive strategy that takes into account farmers' goals, experiences, and degree of knowledge in order to promote the adoption of sustainable technologies.

Prediction of air pollutant concentrations based on the long short-term memory neural network

In recent years, environmental problems caused by air pollutants have received increasing attention. Effective prediction of air pollutant concentrations is an important way to protect the public from harm. Recently, due to extreme climate and social development, the forest fire frequency has increased. During the biomass combustion process caused by forest fires, the content of particulate matter (PM) in the atmosphere increases significantly. However, most existing air pollutant concentration prediction methods do not consider the considerable impact of forest fires, and effective long-term prediction models have not been established to provide early warnings for harmful gases. Therefore, in this paper, we collected a daily air quality data set (aerodynamic diameter smaller than 2.5 µm, PM2.5) for Heilongjiang Province, China, from 2017 to 2023 and A novel Long Short-Term Memory (LSTM) model was proposed to effectively predict the situation of air pollutants. The model could automatically extract information of the effective time step from the historical data set and combine forest fire disturbance and climate data as auxiliary data to improve the model prediction ability. Moreover, we created artificial neural network (ANN) and permissive regression (support vector machine, SVR) models for comparative experiments. The results showed that the precision accuracy of the developed LSTM model is higher. Unlike the other models, the LSTM neural network model could effectively predict the concentration of air pollutants in long-term series. Regarding long-term observation missions (7 days), the proposed model performed well and stably, with R2 reaching over 88%.

Predicting high-intensity resuscitation needs in injured patients in the post-hemostasis phase of care following intervention.

Best resuscitation practices in the posthemostasis phase of care are poorly defined; this phase of care is characterized by a range of physiologic derangements and multiple therapeutic modalities used to address them. Using a cohort of injured patients who required an immediate intervention in the operating room or angiography suite following arrival to the emergency department, we sought to define high-intensity resuscitation (HIR) in this posthemostasis phase of care; we hypothesized that those who would require HIR could be identified, using only data available at intensive care unit (ICU) admission. Clinical data were extracted for consecutive injured patients (2016-2019) admitted to the ICU following an immediate procedure in the operating room or angiography suite. High-intensity resuscitation thresholds were defined as the top decile of blood product (≥3 units) and/or crystalloid (≥4 L) use in the initial 12 hours of ICU care and/or vasoactive medication use between ICU hours 2 and 12. The primary outcome, HIR, was a composite of any of these modalities. Predictive modeling of HIR was performed using logistic regression with predictor variables selected using Least Absolute Shrinkage and Selection Operator (LASSO) estimation. Model was trained using 70% of the cohort and tested on the remaining 30%; model predictive ability was evaluated using area under receiver operator curves. Six hundred five patients were included. Patients were 79% male, young (median age, 39 years), severely injured (median Injury Severity Score, 26), and an approximately 3:2 ratio of blunt to penetrating mechanisms of injury. A total of 215 (36%) required HIR. Predictors selected by LASSO included: shock index, lactate, base deficit, hematocrit, and INR. The area under receiver operator curve for the LASSO-derived HIR prediction model was 0.82. Intensive care unit admission data can identify subsequent HIR in the posthemostasis phase of care. Use of this model may facilitate triage, nursing ratio determination, and resource allocation. Therapeutic/Care Management; Level IV.

Constitutive modeling hot tensile behavior of Ti6Al4V alloy by considering phase transformation and damage mechanisms

The hot tensile and fracture behavior, as well as damage mechanisms of Ti6Al4V alloy are investigated by the hot tension tests and microstructure observation experiments. For comprehensively characterizing the hot tensile behaviors, a developed dislocation density-based constitutive model was established by coupling the flow softening, phase transformation and damage mechanisms. The results show that the flow stresses under the relatively high strain rates (0.01–1 s−1) easily peak at low strains and then decrease with the further straining. However, under the low strain rate (0.001 s−1), the flow stress tends to a continuously increasing trend until fractured. After hot deformation, the fraction of β phase greatly increases compared to that under the initial state. Meanwhile, β-phase fraction increases with the strain rate decreasing or the deformation temperature increasing. The great input thermal energy by the high temperature and the long deformation period under low strain rates promote the α-β phase transformation. Furthermore, the necking diffusion caused by dislocation motion enhances the homogeneous deformation under low strain rate or high temperature. The cavitations preferentially nucleate at the α/β phase interface due to the inconsistent strain. The comparison of the measured and calculated results confirms the excellent prediction ability of the established model in reproducing the hot tensile behavior, β-phase fraction and damage evolution under different processing parameters.

Obtaining strength and ductility synergy for directed energy deposited Ti17 alloys by machine learning

Recently, obtaining a balance of tensile strength and ductility in direct energy deposited (DED) titanium alloy components has been a major concern, which obstructs their further application. Herein, machine learning (ML) methods were applied to find the optimal process window of DEDed titanium alloy parts from a wide range of possible depositing process variables. Four algorithms were used for ML model training and tensile property prediction of DEDed titanium alloy. The results showed that the prediction ability of the XGBoost model was the best. The optimized ‘laser power’-‘scanning rate’ process window produced DEDed titanium alloy specimens with an ultimate tensile strength (UTS) of 1050 MPa and an elongation (E) of 12.5 %. It was demonstrated that ML methods could promote the discovery of the optimal process window, leading to an outstanding synergy of tensile strength and ductility.

External validation of a predictive model for reintubation after cardiac surgery: A retrospective, observational study

Study objectiveExplore validation of a model to predict patients' risk of failing extubation, to help providers make informed, data-driven decisions regarding the optimal timing of extubation. DesignWe performed temporal, geographic, and domain validations of a model for the risk of reintubation after cardiac surgery by assessing its performance on data sets from three academic medical centers, with temporal validation using data from the institution where the model was developed. SettingThree academic medical centers in the United States. PatientsAdult patients arriving in the cardiac intensive care unit with an endotracheal tube in place after cardiac surgery. InterventionsReceiver operating characteristic (ROC) curves and concordance statistics were used as measures of discriminative ability, and calibration curves and Brier scores were used to assess the model's predictive ability. MeasurementsTemporal validation was performed in 1642 patients with a reintubation rate of 4.8%, with the model demonstrating strong discrimination (optimism-corrected c-statistic 0.77) and low predictive error (Brier score 0.044) but poor model precision and recall (Optimal F1 score 0.29). Combined domain and geographic validation were performed in 2041 patients with a reintubation rate of 1.5%. The model displayed solid discriminative ability (optimism-corrected c-statistic = 0.73) and low predictive error (Brier score = 0.0149) but low precision and recall (Optimal F1 score = 0.13). Geographic validation was performed in 2489 patients with a reintubation rate of 1.6%, with the model displaying good discrimination (optimism-corrected c-statistic = 0.71) and predictive error (Brier score = 0.0152) but poor precision and recall (Optimal F1 score = 0.13). Main resultsThe reintubation model displayed strong discriminative ability and low predictive error within each validation cohort. ConclusionsFuture work is needed to explore how to optimize models before local implementation.

A Method for Predicting the Creep Rupture Life of Small-Sample Materials Based on Parametric Models and Machine Learning Models.

In view of the differences in the applicability and prediction ability of different creep rupture life prediction models, we propose a creep rupture life prediction method in this paper. Various time-temperature parametric models, machine learning models, and a new method combining time-temperature parametric models with machine learning models are used to predict the creep rupture life of a small-sample material. The prediction accuracy of each model is quantitatively compared using model evaluation indicators (RMSE, MAPE, R2), and the output values of the most accurate model are used as the output values of the prediction method. The prediction method not only improves the applicability and accuracy of creep rupture life predictions but also quantifies the influence of each input variable on creep rupture life through the machine learning model. A new method is proposed in order to effectively take advantage of both advanced machine learning models and classical time-temperature parametric models. Parametric equations of creep rupture life, stress, and temperature are obtained using different time-temperature parametric models; then, creep rupture life data, obtained via equations under other temperature and stress conditions, are used to expand the training set data of different machine learning models. By expanding the data of different intervals, the problem of the low accuracy of the machine learning model for the small-sample material is solved.

Prediction of landslide step-like displacement using factor preprocessing-based hybrid optimized SVR model in the Three Gorges Reservoir, China

Artificial intelligence (AI) is gaining popularity in landslide displacement modeling due to its strong generalization capacity and ability to accurately represent complex, nonlinear laws. However, few studies have examined the validity of factor selection and the impact of data noise on the model. To address these issues, we employ various analyses, including Pearson correlation coefficient and grey correlation degree analyses to diagnose redundant and highly correlated factors, respectively. Singular spectrum analysis (SSA) is employed for denoising the data, leveraging the benefits of the hankel matrix and minimal parameters. This technique facilitates the decomposition of displacement into trend and periodic terms. The trend term displacement is fitted with a polynomial, and the periodic term displacement is predicted using a support vector regression (SVR) model. The hybrid-optimized SVR by particle swarm optimization-gravitational search algorithm (PSO-GSA) is used to improve the model's performance and prediction ability. We apply the proposed method to the Bazimen and Baishuihe landslides in Zigui County for model validation. The results show that denoising the input factor data reduces the root mean square error (RMSE) and mean absolute error (MAE) and increases the correlation coefficient (R). Furthermore, the PSO-GSA algorithm exhibits excellent generalization ability and fast optimization speed in optimizing parameters. The correlation coefficient increase to 0.99 when using the hybrid-optimized SVR model for displacement prediction.

Estimation of Shear Strength Parameters from Easily-Collected Soil Physical Properties Using Bagging Learning Technique

Shear strength parameters, including cohesion and friction angle, are among the most crucial factors in soil mechanics, playing a pivotal role in the design and construction of engineering projects. This paper aims to estimate these essential soil shear strength parameters using an ensemble learning model. To achieve this, the current study employs the Random Forest (RF) model incorporating various physical parameters of soil, such as density (?), saturation degree (Sr), liquid limit (LL), silt content (SC), clay content (CC) to predict cohesion (c), and friction angle (?). In order to assess the predictive performance of the used model, this research used various metrics, including the mean absolute error (MAE), root mean square error (RMSE), and correlation coefficient (R2), to evaluate the model’s accuracy. The results reveal that RF performs superior predictive capabilities. Furthermore, the proposed model prediction ability was compared to the previous empirical equations. The comparison results indicated that the prediction capability of RF outperforms the previously developed equations.