Standard Prediction Models Research Articles

Predictive Factors for Failed Nonsurgical Management of Long Bone Metastasis and Myeloma.

Understanding the risk factors for failing nonsurgical management of metastatic bone disease is necessary to determine those patients who will benefit from prophylactic stabilization; however, standard predictive models do not include several clinically relevant factors. The primary and secondary objectives of this study were to evaluate comprehensive patient- and disease-related factors as potential predictors of failure of radiation therapy alone for long bone lesions and overall survival in metastatic disease and myeloma. All patients who underwent radiation therapy for long bone metastases at our tertiary care institution from May 2011 to February 2020 were retrospectively reviewed. Of 475 lesions, we excluded those with prophylactic fixation or fracture before radiation therapy, and those <0.5 cm on plain radiographs. Outcomes of the 186 lesions were classified as no progression, progression requiring prophylactic fixation, or progression to pathologic fracture. Blinded radiograph review was done by two orthopaedic oncology surgeons and two musculoskeletal radiologists. Demographic, socioeconomic, lesion, cancer severity, and patient-specific risk factors were identified, and potential predictors were analyzed using backwards stepwise regression. Following radiation therapy, 8.6% lesions underwent prophylactic fixation and 14.0% fractured. Prophylactic fixation was associated with Mirels' score (OR = 1.98, P = 0.025), lesion cortical involvement (OR = 16.96, P = 0.010), and younger patient age (OR = 0.93, P = 0.024). Fracture was associated with lesion cortical involvement (OR = 10.16, P = 0.003) and "low risk" histology (OR = 9.01, P = 0.057). Orthopaedic treatment (either prophylactic surgery or pathologic fracture management) was associated with Mirels' score (OR = 1.62, P = 0.015), lesion cortical involvement (OR = 8.94, P = 0.002), humerus location (OR = 4.19, P = 0.042), and Medicare (OR = 4.12, P = 0.062) or private insurance (OR = 5.69, P = 0.022) compared with Medicaid. ECOG score (OR = 1.28, P = 0.003) was found to be a risk factor for increased mortality after radiotherapy, while "low risk" histology (OR = 0.51, P = 0.029), mixed lesion type (OR = 0.34, P = 0.006), and increased body mass index (OR = 0.95, P = 0.001) were protective factors. Radiograph measurements of cortical involvement were the most clinically relevant for determination of metastatic lesion fracture risk; however, predictors of local failure not addressed in Mirels' score should be considered in clinical decisions about prophylactic fixation. Surgery may be underperformed for histologies commonly considered to be "low risk" for local progression after radiation therapy.

Using phenomic selection to predict hybrid values with NIR spectra measured on the parental lines: proof of concept on maize

Key messagePhenomic selection based on parental spectra can be used to predict GCA and SCA in a sparse factorial design.Prediction approaches such as genomic selection can be game changers in hybrid breeding. They allow predicting the genetic values of hybrids without the need for their physical production. This leads to significant reductions in breeding cycle length, and so to the increase in genetic progress. However, these methods are often underutilized in breeding programs due to the substantial cost involved in genotyping thousands of candidate parental lines annually. To address this limitation, we propose a cost-effective alternative based on phenomic selection, where genotyping of parental lines is replaced by NIR spectroscopy. Standard prediction models are then applied for genomic and phenomic selection, using similarity matrices derived from either genotyping data (genomic selection) or NIR spectral data (phenomic selection). Our hypothesis is that the chemical composition of parental tissues captured by NIRS reflects the genetic similarity between parental lines. We evaluated both strategies using a sparse factorial design, whose hybrids have been phenotyped in a multi-environment trial network, and with NIR spectra acquired on the parental lines on two independent environments. Both genomic and phenomic prediction approaches demonstrated moderate-to-high predictive abilities across various cross-validation scenarios. Our results also showcase the capability of phenomic selection to predict Mendelian sampling. This study serves as a proof of concept that low-cost high-throughput phenomics of parental lines can effectively be used to predict maize hybrids in independent trials. This paves the way for widespread adoption of prediction approaches at the very first stages of hybrid breeding, benefiting both major and orphan species.

Accounting for melting properties in the multi-solid paraffin prediction model for analyzing well production at fields in the Republic of Kazakhstan

The deposition of paraffin on pipeline walls is a serious flow issue. leading to a reduction in oil production due to the decreased flow cross-sectional area. and in severe cases. to complete blockage. This problem also results in increased energy consumption and causes failure of surface equipment due to paraffin plugs. Traditional methods predict paraffin crystallization temperature by considering the impact of temperature on the solubility parameters of individual components in both liquid and solid phases as well as their molar volumes. The goal of this study is to improve the prediction of paraffin deposition in Kazakhstan crude oil by incorporating an evaluation of melting properties into the multi-solid (MS) body prediction model. The process of calculating and adjusting melting properties to enhance the accuracy of the MS model is described in detail. This approach helps to overcome the limitations faced by most existing models when adapting to different types of crude oil. In this study, the focus is on optimizing the multi-solid (MS) body prediction model for paraffin deposition by integrating the melting properties of crude oil components. The improvement of the MS model is crucial because paraffin deposition is a significant issue in Kazakhstan's oil fields, where the wide range of reservoir conditions can cause standard prediction models to fall short. By enhancing the accuracy of melting point calculations, the updated model can better predict the onset of paraffin crystallization and its deposition tendencies under varying pressure and temperature conditions. Keywords: melting properties; multi solid solution; petroleum; solid solution; paraffin deposition

A novel particle size distribution correction method based on image processing and deep learning for coal quality analysis using NIRS-XRF.

The combined application of near-infrared spectroscopy (NIRS) and X-ray fluorescence spectroscopy (XRF) has achieved remarkable results in coal quality analysis by leveraging NIRS's sensitivity to organic compounds and XRF's reliability for inorganic composition. However, variations in particle size distribution negatively affect the diffuse reflectance of NIRS and the fluorescence signal intensities of XRF, leading to decreased accuracy and repeatability in predictions. To address this issue, this study innovatively proposes a particle size correction method that integrates image processing and deep learning. The method first captures micro-images of the coal sample surface using a microscope camera and employs the Segment Anything Model (SAM) for binarization to represent particle size distribution. Subsequently, a Spatial Transformer Network (STN) is applied for geometric correction, followed by feature extraction using a Convolutional Neural Network (CNN) to establish a correlation model between particle size distribution and ash measurement errors. In experiments involving 56 coal samples, including 48at 0.2mm for the standard ash prediction model and 8 within a 0∼1mm range for correction, the results showed significant improvements: standard deviation (SD), mean absolute error (MAE), and root mean square error of prediction (RMSEP) decreased from 0.321%, 0.317%, and 0.335% to 0.229%, 0.225%, and 0.257%, respectively. Using the accuracy of the 0.2mm particle size validation set as a reference, compared to before correction, the errors in these metrics were reduced by 64.06%, 50%, and 60.80%, respectively. This study demonstrates that integrating deep learning and image analysis significantly enhances the repeatability and accuracy of NIRS-XRF measurements, effectively mitigating sub-millimeter particle size effects on spectral detection results and improving model adaptability. This method, through automated particle size distribution analysis and real-time result correction, holds promise for providing essential technical support for the development of online quality detection technologies for conveyor belt materials.

Integration of cost-effective datasets to improve predictability of strategic noise mapping in transport corridors in Delhi city, India.

Assessing exposure to environmental noise levels at transport corridors remains complex in conditions where no standardized noise prediction model is available. In planning and policy implementation for noise control, noise mapping is an important step. In the present study, land use regression model has been developed to predict the environmental noise levels in Delhi city, India, using previously developed approaches along with machine learning techniques, however improved using new datasets. Lday, Lnight, LAeq,24h, and Ldn were modeled at daily resolution by utilizing an annual noise levels dataset from 31 locations in Delhi city. The noise-monitored data was integrated with travel time data, nighttime light data along with common parameters including land use, road networks, and meteorological parameters. The developed LUR models showed good fit with R2 of 0.72 for Lday, 0.55 for Lnight, 0.71 for LAeq,24h, and 0.61 for Ldn, which was further improved up to 0.88 for Lday, 0.79 for Lnight, 0.86 for LAeq,24h, and 0.81 for Ldn by integrating machine learning approaches. The developed models were validated through tenfold cross validation and by comparison to a separate noise level dataset. The average travel time variable was observed to be the most influential predictor of LUR models for Lday and LAeq,24h, which signifies the crucial impact of road traffic congestion on environmental noise levels. The study also analyzed the parametric sensitivity of various infrastructural factors reported in the study, which shall be helpful for planning for smart cities.

Marker weighting improves single-step genomic prediction reliabilities of udder health traits in Nordic Red and Jersey dairy cattle populations

The standard single-step genomic prediction model assumes that all SNP markers explain an equal amount of genetic variance, which, however, may not be true. This is because SNPs are located in or near different genes with different functions. Therefore, it seems logical to consider SNP marker-specific weights when predicting genomic breeding values. We hypothesized that allowing differences in the amount of genetic variance explained by each SNP marker will improve prediction reliability and response to selection. To investigate this hypothesis, we first developed multi-trait standard single-step genomic models based on the current multi-trait random regression evaluation models for udder health traits of the Nordic Red (RDC) and Jersey (JER) dairy cattle populations. The models included 4 clinical mastitis (CM) traits, 3 test-day somatic cell score (SCS) traits, and the conformation traits fore udder attachment and udder depth. In the second step, we investigated the effect of applying different SNP marker weighting scenarios in the single-step genomic prediction models, for which a single-step SNP best linear unbiased prediction model was applied. We investigated the prediction reliability of the different models by forward prediction, where the last 4 years of the data were removed to estimate breeding values for validation candidates. In addition, genetic trends of the pedigree-based estimated breeding values (PEBV) and genomic enhanced breeding values (GEBV) were examined. The data sets for RDC and JER included 6.9 and 1.2 million animals of which 5.6 and 0.9 million cows had records, respectively. The number of genotyped animals was 125,789 and 64,777 for RDC and JER, respectively. Cows had repeated SCS observations but only single observations for all other traits and breeding values for all traits were modeled by one covariance function. This required modeling 12 eigenvalue breeding value coefficients for each cow and developing SNP marker weights for the principal components rather than for the biological traits. We investigated 3 SNP marker weighting scenarios: 1) a nonlinear method similar to BayesA, 2) using the classical formula 2pqû2 that accounts for allele heterozygosity, and 3) applying a mean SNP weight calculated by 2pqû2 for every 20 adjacent SNP markers. Bias, dispersion, and prediction reliability were calculated using PEBV or GEBV from the evaluation based on the full data set on those using the reduced data set. We found that the recent favorable genetic trend in CM and SCS has been accelerated since the introduction of genomic selection. The study also shows that a significant increase in prediction reliability, i.e., 0.74 vs. 0.48 for RDC and 0.72 vs. 0.41 for JER cows for CM, can be achieved with a standard single-step genomic prediction model compared with a pedigree-based prediction model. Almost all scenarios with SNP marker weighting further improved the prediction reliability between 0.5% and 12.7%. The highest improvement was achieved by weighing the SNP markers based on the 2pqû2 formula.

Development of a Multi‐Scale Meteorological Large‐Eddy Simulation Model for Urban Thermal Environmental Studies: The “City‐LES” Model Version 2.0

AbstractTo bridge the gaps between meteorological large‐eddy simulation (LES) models and computational fluid dynamics (CFD) models for microscale urban climate simulations, the present study has developed a meteorological LES model for urban areas. This model simulates urban climates across both mesoscale (city scale) and microscale (city‐block scale). The paper offers an overview of this LES model, which distinguishes itself from standard numerical weather prediction models by resolving buildings and trees at the microscale simulations. It also differs from standard CFD models by accounting for atmospheric stratification and physical processes. Noteworthy features of this model include: (a) the calculation of long‐ and short‐wave radiations in three dimensions, incorporating multiple reflections within urban canopy layers using the radiosity method, and accounting for building and tree shadows in the simulations; (b) the provision of various heat stress indices (Universal Thermal Climate Index, Wet Bulb Globe Temperature, MRT, THI); (c) the assessment of the efficacy of heat stress mitigation measures such as dry‐mist spraying, roadside trees, cool pavements, and green/cool roofs strategies; (d) the capability to run on supercomputers, with the code parallelized in a three‐dimensional manner, and the model can also run on a graphics processing unit cluster. Following the introduction of this model, the study confirms its basic performance through various numerical experiments, including simulations of thermals in the convective boundary layer, coherent structure of turbulence over urban canopy, and thermal environment and heat stress indices in urban districts. The model developed in this study is intended to serve as a community tool for addressing both fundamental and applied studies in urban climatology.

Optimization‐based pore network modeling approach for determination of hydraulic conductivity function of granular soils

AbstractA wide range of applications of unsaturated hydraulic conductivity is well known in geotechnical, hydrological, and agricultural engineering fields. The standard prediction models for hydraulic conductivity function overlook the complexity of soil pore structure and employ a simplistic approach based on the bundle of capillary tubes. This study proposes an alternative approach employing pore network models calibrated to match soil water retention data to predict the hysteretic hydraulic conductivity function of granular soils. A novel approach to constructing a multidirectional pore network built on an irregular lattice with variable coordination numbers is presented for the realistic representation of soil voids. The geometric and topological parameters of the pore network model are optimized using the genetic algorithm, and adequate pore‐scale processes (piston‐like advance and corner flow during drainage and piston‐like advance, pore body filling, and snap‐off during imbibition) are modeled to get reasonable predictions of hysteretic hydraulic conductivity functions over the entire suction range of granular soils. Comparisons between the pore network model results, standard physically based models, and measured data for a variety of granular soils show that the proposed pore network has a superior performance over other models and compares favorably to the experimental data.

A data-centric machine learning approach to improve prediction of glioma grades using low-imbalance TCGA data

Accurate prediction and grading of gliomas play a crucial role in evaluating brain tumor progression, assessing overall prognosis, and treatment planning. In addition to neuroimaging techniques, identifying molecular biomarkers that can guide the diagnosis, prognosis and prediction of the response to therapy has aroused the interest of researchers in their use together with machine learning and deep learning models. Most of the research in this field has been model-centric, meaning it has been based on finding better performing algorithms. However, in practice, improving data quality can result in a better model. This study investigates a data-centric machine learning approach to determine their potential benefits in predicting glioma grades. We report six performance metrics to provide a complete picture of model performance. Experimental results indicate that standardization and oversizing the minority class increase the prediction performance of four popular machine learning models and two classifier ensembles applied on a low-imbalanced data set consisting of clinical factors and molecular biomarkers. The experiments also show that the two classifier ensembles significantly outperform three of the four standard prediction models. Furthermore, we conduct a comprehensive descriptive analysis of the glioma data set to identify relevant statistical characteristics and discover the most informative attributes using four feature ranking algorithms.

CGNSDE: Conditional Gaussian neural stochastic differential equation for modeling complex systems and data assimilation

A new knowledge-based and machine learning hybrid modeling approach, called conditional Gaussian neural stochastic differential equation (CGNSDE), is developed to facilitate modeling complex dynamical systems and implementing analytic formulae of the associated data assimilation (DA). In contrast to the standard neural network predictive models, the CGNSDE is designed to effectively tackle both forward prediction tasks and inverse state estimation problems. The CGNSDE starts by exploiting a systematic causal inference via information theory to build a simple knowledge-based nonlinear model that nevertheless captures as much explainable physics as possible. Then, neural networks are supplemented to the knowledge-based model in a specific way, which not only characterizes the remaining features that are challenging to model with simple forms but also advances the use of analytic formulae to efficiently compute the nonlinear DA solution. These analytic formulae are used as an additional computationally affordable loss to train the neural networks that directly improve the DA accuracy. This DA loss function promotes the CGNSDE to capture the interactions between state variables and thus advances its modeling skills. With the DA loss, the CGNSDE is more capable of estimating extreme events and quantifying the associated uncertainty. Furthermore, crucial physical properties in many complex systems, such as the translate-invariant local dependence of state variables, can significantly simplify the neural network structures and facilitate the CGNSDE to be applied to high-dimensional systems. Numerical experiments based on chaotic systems with intermittency and strong non-Gaussian features indicate that the CGNSDE outperforms knowledge-based regression models, and the DA loss further enhances the modeling skills of the CGNSDE.

Construction and validation of a risk prediction model for postoperative delirium in patients with off‑pump coronary artery bypass grafting.

Compared with cardiopulmonary bypass surgery, off-pump coronary artery bypass grafting (OPCABG) reduces trauma to the body. However, there is still a risk of neurological complications, including postoperative delirium (POD). To date, few studies have been conducted on the risk of POD in OPCABG patients, and no standardized prediction model has been established. Thus, this study sought to analyze the factors influencing POD in OPCABG patients and to construct a risk prediction model. A total of 1,258 patients with OPCABG were enrolled and divided into the training set for model construction (944 cases) and the test set for model validation (314 cases). A risk prediction model for POD in OPCABG patients was established by least absolute shrinkage and selection operator (LASSO) regression and multivariate logistic regression, and a nomogram was drawn. The discrimination and calibration degree of the model was evaluated by the receiver operator characteristic (ROC) curve and calibration curve. Eight variables [i.e., age, tissue oxygen saturation, mean arterial pressure (MAP), carotid stenosis, the anterior-posterior diameter of the aortic sinus, ventricular septum thickness, left ventricular ejection fraction (LVEF), and Mini-Mental State Examination (MMSE) scores] were screen out by the LASSO regression and multivariate logistic regression, and the model was constructed. The area under the ROC curve of the training set was 0.702 [95% confidence interval (CI): 0.662-0.743], and that of the test set was 0.658 (95% CI: 0.585-0.730). The results of the Hosmer-Lemeshow goodness-of-fit test showed that the predicted POD risk of OPCABG patients in the training and test sets was consistent with the actual POD risk (χ2=5.154, P=0.74). The occurrence of POD in OPCABG patients is related to age, tissue oxygen saturation, MAP, carotid artery stenosis, the anterior-posterior diameter of aortic sinus, ventricular septal thickness, LVEF, and MMSE scores. The prediction model constructed with the above variables had high predictive performance, and thus may be helpful in the early identification of such patients.

Application of fusion property calculations for Kazakhstani oil samples

This study thoroughly investigates the melting properties and solid-state transition temperatures of crude oil from Kazakhstan's oil field, which is critical for predicting wax precipitation. Utilizing advanced laboratory equipment such as gas chromatographs, differential scanning calorimeters (DSC), and pour point testers, we obtained comprehensive data on these parameters. Based on the results, we developed new correlations for the melting point and solid-state transition temperature of crude oil, which demonstrate closer alignment with the reference values for various hydrocarbons' melting points compared to standard prediction models. These findings provide more accurate models for wax deposition, essential for efficient crude oil transportation and processing. The results indicated that implementing these new correlations could lead to significant cost savings by reducing pipeline blockages and maintenance requirements, thereby improving overall operational efficiency.

Optimization of garment sewing operation standard minute value prediction using an IPSO-BP neural network

Abstract Standard minute value serves as a pivotal metric guiding the arrangement and balancing of production cycles in clothing production lines, and plays a crucial role in cost pricing and production order arrangement for clothing products. Given the complexity of the garment sewing process, ten influencing factors including fabric weight, fabric thickness, fabric density, stitching length, stitching shapes, cut pieces numbers, notch numbers, sewing technologies, sewing machine, and auxiliary accessories were identified. Upon this foundation, a standard sewing time prediction model, Improved particle swarm optimization - Back-propagation neural network (IPSO-BP), was proposed, focusing on non-quantitative factors. The IPSO-BP model was trained using actual sewing data from a women’s clothing production company. Compared to the unoptimized BP neural network, the IPSO-BP model demonstrated significant advantages in terms of convergence speed and prediction accuracy. Therefore, the IPSO-BP model proposed in this study holds promise for predicting standard sewing hours effectively.

Just-in-Time crash prediction for mobile apps

Just-In-Time (JIT) defect prediction aims to identify defects early, at commit time. Hence, developers can take precautions to avoid defects when the code changes are still fresh in their minds. However, the utility of JIT defect prediction has not been investigated in relation to crashes of mobile apps. We therefore conducted a multi-case study employing both quantitative and qualitative analysis. In the quantitative analysis, we used machine learning techniques for prediction. We collected 113 reliability-related metrics for about 30,000 commits from 14 Android apps and selected 14 important metrics for prediction. We found that both standard JIT metrics and static analysis warnings are important for JIT prediction of mobile app crashes. We further optimized prediction performance, comparing seven state-of-the-art defect prediction techniques with hyperparameter optimization. Our results showed that Random Forest is the best performing model with an AUC-ROC of 0.83. In our qualitative analysis, we manually analysed a sample of 642 commits and identified different types of changes that are common in crash-inducing commits. We explored whether different aspects of changes can be used as metrics in JIT models to improve prediction performance. We found these metrics improve the prediction performance significantly. Hence, we suggest considering static analysis warnings and Android-specific metrics to adapt standard JIT defect prediction models for a mobile context to predict crashes. Finally, we provide recommendations to bridge the gap between research and practice and point to opportunities for future research.

(249) - Machine Learning Outperforms Standard Prediction Models for 3 Year Prognosis Following Lung Transplantation

(249) - Machine Learning Outperforms Standard Prediction Models for 3 Year Prognosis Following Lung Transplantation

Benchmarking machine learning-based real-time respiratory signal predictors in 4D SBRT.

Stereotactic body radiotherapy of thoracic and abdominal tumors has to account for respiratory intrafractional tumor motion. Commonly, an external breathing signal is continuously acquired that serves as a surrogate of the tumor motion and forms the basis of strategies like breathing-guided imaging and gated dose delivery. However, due to inherent system latencies, there exists a temporal lag between the acquired respiratory signal and the system response. Respiratory signal prediction models aim to compensate for the time delays and to improve imaging and dosedelivery. The present study explores and compares six state-of-the-art machine and deep learning-based prediction models, focusing on real-time and real-world applicability. All models and data are provided as open source and data to ensure reproducibility of the results and fosterreuse. The study was based on 2502 breathing signals ( h) acquired during clinical routine, split into independent training (50%), validation (20%), and test sets (30%). Input signal values were sampled from noisy signals, and the target signal values were selected from corresponding denoised signals. A standard linear prediction model (Linear), two state-of-the-art models in general univariate signal prediction (Dlinear, Xgboost), and three deep learning models (Lstm, Trans-Enc, Trans-TSF) were chosen. The prediction performance was evaluated for three different prediction horizons (480, 680, and 920ms). Moreover, the robustness of the different models when applied to atypical, that is, out-of-distribution (OOD) signals, was analyzed. The Lstm model achieved the lowest normalized root mean square error for all prediction horizons. The prediction errors only slightly increased for longer horizons. However, a substantial spread of the error values across the test signals was observed. Compared to typical, that is, in-distribution test signals, the prediction accuracy of all models decreased when applied to OOD signals. The more complex deep learning models Lstm and Trans-Enc showed the least performance loss, while the performance of simpler models like Linear dropped the most. Except for Trans-Enc, inference times for the different models allowed for real-time application. The application of the Lstm model achieved the lowest prediction errors. Simpler prediction filters suffer from limited signal history access, resulting in a drop in performance for OODsignals.

Development and validation of an individualized and weighted Myeloma Prognostic Score System (MPSS) in patients with newly diagnosed multiple myeloma.

Current standard predictive models of disease risk do not adequately account for the heterogeneity of survival outcomes in patients with new-diagnosed multiple myeloma (NDMM). In this retrospective, multicohort study, we collected clinical and genetic data from 1792 NDMM patients and identified the prognostic impact of all features. Using the top-ranked predictive features, a weighted Myeloma Prognostic Score System (MPSS) risk model was formulated and validated to predict overall survival (OS). In the training cohort, elevated lactate dehydrogenase level (LDH), International Staging System (ISS) Stage III, thrombocytopenia, and cumulative high-risk cytogenetic aberration (HRA) numbers were found to have independent prognostic significance. Each risk factor was defined as its weighted value respectively according to their hazard ratio for OS (thrombocytopenia 2, elevated LDH 1, ISS III 2, one HRA 1, and ≥2 HRA 2, points). Patients were further stratified into four risk groups: MPSS I (22.5%, 0 points), II (17.6%, 1 points), III (38.6%, 2-3 points), and IV (21.3%, 4-7 points). MPSS risk stratification showed optimal discrimination, as well as calibration, of four risk groups with median OS of 91.0, 69.8, 45.0, and 28.0 months, for patients in MPSS I to IV groups (p < .001), respectively. Importantly, the MPSS model retained its prognostic value in the internal validation cohort and an independent external validation cohort, and exhibited significant risk distribution compared with conventional prognostic models (R-ISS, R2-ISS, and MASS). Utilization of the MPSS model in clinical practice could improve risk estimation in NDMM patients, thus prompting individualized treatment strategies.

Optimal profit-making strategies in stock market with algorithmic trading

&lt;p&gt;Machine learning (ML) techniques are being increasingly applied to financial markets for analyzing trends and predicting stock prices. In this study, we compared the price prediction and profit-making performance of various ML algorithms embedded into stock trading strategies. The dataset comprised daily data from the CSI 300 Index of the China stock market spanning approximately 17 years (2006–2023). We incorporated investor sentiment indicators and relevant financial elements as features. Our trained models included support vector machines (SVMs), logistic regression, and random forest. The results show that the SVM model outperforms the others, achieving an impressive 60.52% excess return in backtesting. Furthermore, our research compared standard prediction models (such as LASSO and LSTM) with the proposed approach, providing valuable insights for users selecting ML algorithms in quantitative trading strategies. Ultimately, this work serves as a foundation for informed algorithm choice in future financial applications.&lt;/p&gt;

Development and Internal Validation of Machine Learning Models to Predict Mortality and Disability After Mechanical Thrombectomy for Acute Anterior Circulation Large Vessel Occlusion

Mechanical thrombectomy (MT) improves outcomes in patients with LVO but many still experience mortality or severe disability. We sought to develop machine learning (ML) models that predict 90-day outcomes after MT for LVO. Consecutive patients who underwent MT for LVO between 2015-2021 at a Comprehensive Stroke Center were reviewed. Outcomes included 90-day favorable functional status (mRS 0-2), severe disability (mRS 4-6), and mortality. ML models were trained for each outcome using prethrombectomy data (pre) and with thrombectomy data (post). Three hundred and fifty seven patients met the inclusion criteria. After model screening and hyperparameter tuning the top performing ML model for each outcome and timepoint was random forest (RF). Using only prethrombectomy features, the AUCs for the RFpre models were 0.73 (95% CI 0.62-0.85) for favorable functional status, 0.77 (95% CI 0.65-0.86) for severe disability, and 0.78 (95% CI 0.64-0.88) for mortality. All of these were better than a standard statistical model except for favorable functional status. Each RF model outperformed Pre, SPAN-100, THRIVE, and HIAT scores (P<0.0001 for all). The most predictive features were premorbid mRS, age, and NIHSS. Incorporating MT data, the AUCs for the RFpost models were 0.80 (95% CI 0.67-0.90) for favorable functional status, 0.82 (95% CI 0.69-0.91) for severe disability, and 0.71 (95% CI 0.55-0.84) for mortality. RF models accurately predicted 90-day outcomes after MT and performed better than standard statistical and clinical prediction models.

Shrinkage model for waste clay brick-based geopolymer concrete

This study investigated the shrinkage behaviour of geopolymer concrete (GPC) under various curing environments. One-part GPC was prepared by using waste clay bricks (WCB), fly ash and slag as precursors and anhydrous sodium silicate as the sole activator. The samples were subjected to initial curing in three different environments: standard curing in lime-saturated water, unsealed ambient curing, and sealed ambient curing. Irrespective of the curing environment, GPC showed a higher shrinkage than conventional concrete with ordinary Portland cement. Compared to sealed and unsealed curing, standard curing was able to reduce the shrinkage in GPC by more than 22%, considering the 28-day shrinkage strain. However, this curing environment was not suitable in terms of strength development. For WCB-based GPC, the sealed curing improved the strength gain, and also reduced the shrinkage compared to unsealed curing by more than 50%. The existing standard prediction models for Portland cement concrete significantly underestimated the shrinkage of GPC. This study proposed a new model with higher accuracy to predict the shrinkage of GPC. The model used a 28-day shrinkage value to estimate the long-term behaviour, and has the flexibility to adjust its coefficient based on the exposure environment.