Greater Predictive Power Research Articles

PREDICTION BASED ON MACHINE LEARNING OF TOOTH SENSITIVITY FOR IN-OFFICE DENTAL BLEACHING

To develop a supervised machine learning model to predict the occurrence and intensity of tooth sensitivity (TS) in patients undergoing in-office dental bleaching testing various algorithm models. Retrospective data from 458 patients were analyzed, including variables such as the occurrence and intensity of TS, basal tooth color, bleaching material characteristics (concentration and pH), intervention details (number and duration of applications), and patient age. Classification and regression models were evaluated using 5-fold cross-validation and assessed based on various performance parameters. For the predictive classification task (occurrence of TS), the developed models achieved a maximum area under the receiver operating characteristic curve (AUC) of 0.76 [0.62-0.88] on the test data, with an F1-score of 0.80 [0.71-0.87]. In cross-validation, the highest AUC reached 0.86 [0.84-0.88], and the highest F1-score was 0.78 [0.75-0.83]. For predicting TS intensity, the regression models demonstrated a minimum mean absolute error (MAE) of 1.76 [1.45-2.06] and a root mean square error (RMSE) of 2.38 [2.06-2.69] on the test set. During cross-validation, the lowest MAE was 1.84 [1.67-2.03], with an RMSE of 2.39 [2.20-2.58]. The supervised machine learning model for estimating the occurrence and intensity of TS in patients undergoing in-office bleaching demonstrated good predictive power. The Gradient Boosting Classifier and Support Vector Machine Regressor algorithms stood out as having the greatest predictive power among those tested. These models can serve as valuable tools for anticipating tooth sensitivity in this patient population, facilitating better post-treatment management and control.

Inferring the locomotor ecology of two of the oldest fossil squirrels: influence of operationalization, trait, body size and machine learning method.

Correlations between morphology and lifestyle of extant taxa are useful for predicting lifestyles of extinct relatives. Here, we infer the locomotor behaviour of Palaeosciurus goti from the middle Oligocene and Palaeosciurus feignouxi from the lower Miocene of France using their femoral morphology and different machine learning methods. We used two ways to operationalize morphology, in the form of a geometric morphometric shape dataset and a multivariate dataset of 11 femoral traits. The predictive models were built and tested using more than half (180) of the extant species of squirrel relatives. Both traditional models such as linear discriminant analysis and more sophisticated models like neural networks had the greatest predictive power. However, the predictive power also depended on the operationalization and the femoral traits used to build the model. We also found that predictive power tended to improve with increasing body size. Contrary to previous suggestions, the older species, P. goti, was most likely arboreal, whereas P. feignouxi was more likely terrestrial. This provides further evidence that arboreality was already the most common locomotor ecology among the earliest squirrels, while a predominantly terrestrial locomotor behaviour evolved shortly afterwards, before the vast establishment of grasslands in Europe.

Diagnosis of human envenoming by terrestrial venomous animals: Routine, advances, and perspectives

Despite the development of new and advanced diagnostic approaches, monitoring the clinical evolution of accidents caused by venomous animals is still a challenge for science. In this review, we present the state of the art of laboratory tests that are routinely used for the diagnosis and monitoring of envenomings by venomous animals, as well as the use of new tools for more accurate and specific diagnoses. While a comprehensive range of tools is outlined, comprising hematological, biochemical, immunoassays, and diagnostic imaging tools, it is important to acknowledge their limitations in predicting the onset of clinical complications, since they provide an overview of organic damage after its development. Thus, the need for discovery, validation, and use of biomarkers that have greater predictive power, sensitivity and specificity is evident. This will help in the diagnosis, monitoring, and treatment of patients envenomated by venomous animals, consequently reducing the global burden of morbidity and mortality.

Abstract P313: Diastolic to Systolic Arterial stiffening has Greater Prognostic Power than Diastolic Stiffness in 20-60 Years Old Patients Undergoing 24-Hour Ambulatory Blood Pressure Monitoring

The 24-hour ambulatory pulse pressure (PP), can be expressed as a sum of two ‘PP components’: ‘elastic’ PP (elPP) and ‘stiffening’ PP (stPP) associated, respectively, with arterial stiffness at the diastolic pressure (somewhat equivalent to pulse wave velocity being measured in the diastole), and the relative stiffness change from the diastolic- to the systolic pressure. Here we assess and compare the hazard ratio (HR) of elPP and stPP for outcomes in a large cohort of young and middle-age patients. We analyzed adequate quality 24h ambulatory blood pressure monitoring records from the Spanish Ambulatory Blood Pressure Registry. For standard deviation(SD), K=SD(SBP)/SD(DBP), elPP=PPxln(K)/(K-1) and stPP=PP-elPP. HRs were determined for age 20-40, 41-50, and 51-60 years, using Cox regression models applied to standardized elPP & stPP (combined), adjusted for age, sex, BMI, smoking status, diabetes, dyslipidemia, mean DBP, pulse rate, SD(SBP), antihypertensive treatment, and previous cardiovascular disease (CVD). The endpoints were total mortality, and death from CVD, or coronary heart disease (CHD). Of the 24,260 patients mean age (SD) 48.0(9.0) years, 34% male] 845 deaths occurred during period of 9.6(2.2) years. The Table shows that mean elPP slightly decreased from younger to older age, while stPP strongly increased. At age 41-60 years the adjusted HRs of stPP appeared significant and greater than those of elPP for almost all cases. The most contributing adjustors in the Cox regression were smoking, for all age classes, and sex for age over 40 years. Thus, the stiffening component may have greater predictive power than the elastic component for total mortality, and CVD&CHD death, in middle-age patients.

Mining educational big data to develop an early alert dynamic model of academically at-risk students: A proof of concept

ABSTRACT Educational big data has great predictive power to identify academically at-risk students. Based on fully mining educational big data, this study developed an early alert dynamic model of academically at-risk students to confirm and extend this proposition. An empirical test was carried out to validate the reliability and validity of the dynamic model by using data analysis methods. The results proved that the developed model could precisely identify academically at-risk students, scientifically assess their academic risk status, and dynamically provide personalised interventions. The overall prediction success rate of the developed model was 80.8%. In addition, application reflections were discussed to shed light on future implementation.

Explainable Machine Learning Approach to Prediction of Prolonged Intesive Care Unit Stay in Adult Spinal Deformity Patients: Machine Learning Outperforms Logistic Regression.

Retrospective cohort study. Prolonged ICU stay is a driver of higher costs and inferior outcomes in Adult Spinal Deformity (ASD) patients. Machine learning (ML) models have recently been seen as a viable method of predicting pre-operative risk but are often 'black boxes' that do not fully explain the decision-making process. This study aims to demonstrate ML can achieve similar or greater predictive power as traditional statistical methods and follows traditional clinical decision-making processes. Five ML models (Decision Tree, Random Forest, Support Vector Classifier, GradBoost, and a CNN) were trained on data collected from a large urban academic center to predict whether prolonged ICU stay would be required post-operatively. 535 patients who underwent posterior fusion or combined fusion for treatment of ASD were included in each model with a 70-20-10 train-test-validation split. Further analysis was performed using Shapley Additive Explanation (SHAP) values to provide insight into each model's decision-making process. The model's Area Under the Receiver Operating Curve (AUROC) ranged from 0.67 to 0.83. The Random Forest model achieved the highest score. The model considered length of surgery, complications, and estimated blood loss to be the greatest predictors of prolonged ICU stay based on SHAP values. We developed a ML model that was able to predict whether prolonged ICU stay was required in ASD patients. Further SHAP analysis demonstrated our model aligned with traditional clinical thinking. Thus, ML models have strong potential to assist with risk stratification and more effective and cost-efficient care.

Three-dimensional geometric morphometric analyses of humerus ecomorphology: New perspectives for paleohabitat reconstruction in carnivorans and ungulates.

Long bone ecomorphology has proven effective for paleohabitat reconstructions across a wide range of mammalian clades. Still, there is no comprehensive framework to allow interpretation of long bone morphological variation within and between different monophyletic groups. Here, we investigated the use of humerus morphometry to classify living members of the orders Carnivora and ungulates based on their preferred habitats. Using geometric morphometrics, we extracted three different kinds of humerus shape data describing interspecific variation with and without accounting for evolutionary allometry and phylogenetic signal. The traditional a priori categorization of species in open, mixed, and closed habitats was employed in combination with selected subsets of shape variables to identify the best-predictive models for habitat adaptation. These were identified based on the statistical performance of phylogenetic and non-phylogenetic discriminant analyses and then applied to predict habitats on a subsample of fossil species. Size-free shape data combined with phylogenetic discriminant analyses showed the highest rate of accuracy in habitat classification for a combined sample of carnivorans and ungulates. Conversely, when the two groups were investigated separately, traditional shape data analyzed with phylogenetic discriminant function analyses provided models with the greatest predictive power. By combining carnivorans and ungulates within the same methodological framework we identified common adaptive features in closed habitat-adapted species that show compressed epiphyses, while open habitat-adapted species have expanded epiphyses. These morphologies evolved to allow significant degree of direction switches during locomotion in closed habitats compared to open habitat-adapted species whose forelimb joints evolved to stabilize articulations for increasing speed.

Advancing credit risk modelling with Machine Learning: A comprehensive review of the state-of-the-art

Ensuring financial stability necessitates responsible credit granting, so lending institutions maintain sufficient regulatory capital to withstand losses from defaults. Classification methods like Logistic Regression (LR) have long been the standard for estimating default likelihood, but they struggle with increasing operational demands and regulatory standards. Advances in Machine Learning (ML) now allow models to handle larger datasets with greater predictive power. Algorithms such as Decision Trees (DT), Support Vector Machines (SVM), and Neural Networks (NN), though not readily transferable to practical applications, have the potential to enhance the credit risk modelling process. While much of the literature on ML in credit risk focuses on sophisticated, experimental algorithms not yet ready for commercial use, few address the steps and nuances of model development. Our article fills this gap by systematically reviewing the literature on ML in consumer credit risk modelling, highlighting the current state of scientific knowledge. We analyze the frequent steps involved and demonstrate how ML's strengths can be leveraged throughout them. From pre-processing steps like data cleaning, transformation, and reduction, to variable selection and parameter optimization, our findings indicate that ML is becoming increasingly important in financial risk management. However, standardized modelling procedures are needed before complex ML classifiers can be commercially deployed. Meanwhile, although relatively interpretable ML models built around DTs and Random Forests (RF) do showcase increased performance in popular credit risk datasets, it is clear that the potential to benefit from ML spans further, with complex models enabling increased performance by aiding simpler, explainable classification algorithms.

Identification of lipid metabolism-related gene signature in the bone marrow microenvironment of multiple myelomas through deep analysis of transcriptomic data

Dysregulated lipid metabolism in the bone marrow microenvironment (BMM) plays a vital role in multiple myeloma (MM) development, progression, and drug resistance. However, the exact mechanism by which lipid metabolism impacts the BMM, promotes tumorigenesis, and triggers drug resistance remains to be fully elucidated.By analyzing the bulk sequencing and single-cell sequencing data of MM patients, we identified lipid metabolism-related genes differential expression significantly associated with MM prognosis, referred to as LMRPgenes. Using a cohort of ten machine learning algorithms and 117 combinations, LMRPgenes predictive models were constructed. Further exploration of the effects of the model risk score (RS) on the survival status, immune status of patients with BMM, and response to immunotherapy was conducted. The study also facilitated the identification of personalized therapeutic strategies targeting specified risk categories within patient cohorts.Analysis of the scRNA-seq data revealed increased lipid metabolism-related gene enrichment scores (LMESs) in erythroblasts and progenitor, malignant, and Tprolif cells but decreased LMESs in lymphocytes. LMESs were also strongly correlated with most of the 50 hallmark pathways within these cell populations. An elevated malignant cell ratio and reduced lymphocytes were observed in the high LMES group. Moreover, the LMRPgenes predictive model, consisting of 14 genes, showed great predictive power. The risk score emerged as an independent indicator of poor outcomes. Inverse relationships between the RS and immune status were noted, and a high RS was associated with impaired immunotherapy responses. Drug sensitivity assays indicated the effectiveness of bortezomib, buparlisib, dinaciclib, staurosporine, rapamycin, and MST-312 in the high-RS group, suggesting their potential for treating patients with high-RS values and poor response to immunotherapy. Ultimately, upon verification via qRT-PCR, we observed a significant upregulation of ACBD6 in NDMM group compared to the control group.Our research enhances the knowledge base regarding the association between lipid metabolism-related genes (LMRGs) and the BMM in MM patients, offering substantive insights into the mechanistic effects of the BMM mediated by LMRGs.Graphical abstract

The power of machine learning methods to predict crowdfunding success: Accounting for complex relationships efficiently

The objective of this paper is to both demonstrate and explain the power of machine learning (ML) methods to predict crowdfunding success. The first step to achieve this objective is to compare the predictive performance of four ML methods (boosted trees, random forest, Shallow Neural Networks and Deep Neural Networks) to standard binary logit estimation using a dataset of more than 108,223 Kickstarter projects. The data used for the comparison is a set of categorical and continuous variables for each project that can be analyzed using all 5 modelling methods. The second step to achieve the objective is to show that the presence of complex relationships between the explanatory variables and outcome variable explains the power of ML methods. These relationships can be categorized as threshold-type (when a certain level of a explanatory variable is need for an effect to occur), Goldilocks-type (where a very specific level of an explanatory variable is needed for an effect to occur) and interactions (where the effect of one explanatory variable on the outcome variable is moderated by the levels of other explanatory variables). Our estimations show that these relationships are embedded in the context of crowdfunding. Because crowdfunding is a marketing activity, factors like attention, saturation, and synergy between levers are pervasive. It is because machine learning has the ability to account for these relationships (without knowing where they are in advance) that these methods are both better predictors and better managerial tools than standard binary logit models. We further show that even greater prediction power is unleashed by activating the text-analysis capabilities of ML methods. However, even when this is not possible, the value realized by employing ML approaches to predict crowdfunding success is substantial.

Potential distribution and ecological impacts of Acmella radicans (Jacquin) R.K. Jansen (a new Yunnan invasive species record) in China

Background Acmella radicans(Jacquin) R.K. Jansen is a new invasive species record for Yunnan Province, China. Native to Central America, it has also been recently recorded invading other parts of Asia. To prevent this weed from becoming a serious issue, an assessment of its ecological impacts and potential distribution is needed. We predicted the potential distribution of A. radicans in China using the MaxEnt model and its ecological impacts on local plant communities and soil nutrients were explored.ResultsSimulated training using model parameters produced an area under curve value of 0.974, providing a high degree of confidence in model predictions. Environmental variables with the greatest predictive power were precipitation of wettest month, isothermality, topsoil TEB (total exchangeable bases), and precipitation seasonality, with a cumulative contribution of more than 72.70% and a cumulative permutation importance of more than 69.20%. The predicted potential suitable area of A. radicans in China is concentrated in the southern region. Projected areas of A. radicans ranked as high and moderately suitable comprised 5425 and 26,338 km2, accounting for 0.06 and 0.27% of the Chinese mainland area, respectively. Over the 5 years of monitoring, the population density of A. radicans increased while at the same time the population density and importance values of most other plant species declined markedly. Community species richness, diversity, and evenness values significantly declined. Soil organic matter, total N, total P, available N, and available P concentrations decreased significantly with increasing plant cover of A. radicans, whereas pH, total K and available K increased.ConclusionOur study was the first to show that A. radicans is predicted to expand its range in China and may profoundly affect plant communities, species diversity, and the soil environment. Early warning and monitoring of A. radicans must be pursued with greater vigilance in southern China to prevent its further spread.

Development and validation of nomograms for predicting the prognosis of early and late recurrence of advanced gastric cancer after radical surgery based on post-recurrence survival.

In this study, we aimed to explore the risk factors influencing post-recurrence survival (PRS) of early recurrence (ER) and late recurrence (LR) in stage advanced gastric cancer (AGC) patients after radical surgery, respectively, and to develop predictive models in turn. Medical records of 192 AGC patients who recurred after radical gastrectomy were retrospectively reviewed. They were randomly divided into the training and validation set at a ratio of 2:1. Nomograms were built based on risk factors influencing PRS of ER and LR explored by Cox regression analyses, respectively. Concordance index (C-index) values and calibration curves were used to evaluate predictive power of nomograms. Body mass index < 18.5 kg/m2, prealbumin level < 70.1 mg/L, positive lymph nodes ratio ≥ 0.486 and palliative treatment after recurrence were independent risk factors for the prognosis of ER. In contrast, prealbumin level < 170.1 mg/L, CEA ≥ 18.32 μg/L, tumor diameter ≥ 5.5 cm and palliative treatment after recurrence were independent risk factors for the prognosis of LR. The C-index values were 0.801 and 0.772 for ER and LR in the training set, respectively. The calibration curves of validation set showed a C-index value of 0.744 and 0.676 for ER and LR, respectively. Nomograms which were constructed to predict the prognosis of ER and LR of AGC after surgery showed great predictive power and could provide reference for clinicians' treatment strategies to some extent.

Time to maximum amplitude of thromboelastography can predict mortality in patients with severe COVID-19: a retrospective observational study.

To predict mortality in severe patients with COVID-19 at admission to the intensive care unit (ICU) using thromboelastography (TEG). This retrospective, two-center, observational study involved 87 patients with PCR-and chest CT-confirmed severe COVID-19 who were admitted to at Wuhan Huoshenshan Hospital and the 908th Hospital of Chinese PLA Logistic Support Force between February 2020 and February 2023. Clinic demographics, laboratory results, and outcomes were compared between those who survived and those who died during hospitalization. Thromboelastography showed that of the 87 patients, 14 were in a hypercoagulable state, 25 were in a hypocoagulable state, and 48 were normal, based on the time to maximum amplitude (TMA). Patients who died showed significantly lower α angle, but significantly longer R-time, K-time and TMA than patients who survived. Random forest selection showed that K-time, TMA, prothrombin time (PT), international normalized ratio (INR), D-dimer, C-reactive protein (CRP), aspartate aminotransferase (AST), and total bilirubin (Tbil) were significant predictors. Multivariate logistic regression identified that TMA and CRP were independently associated with mortality. TMA had a greater predictive power than CRP levels based on time-dependent AUCs. Patients with TMA ≥ 26.4 min were at significantly higher risk of mortality (hazard ratio 3.99, 95% Confidence Interval, 1.92-8.27, p < 0.01). TMA ≥26.4 min at admission to ICU may be an independent predictor of in-hospital mortality for patients with severe COVID-19.

Identification of PANoptosis-related biomarkers and analysis of prognostic values in head and neck squamous cell carcinoma

PANoptosis plays a crucial role in cancer initiation and progression. However, the roles of PANoptosis-related genes (PARGs) in the prognosis and immune landscape of head and neck squamous cell carcinoma (HNSCC) remain unclear. Integrated bioinformatics analyses based on the data of HNSCC patients in the TCGA database were conducted. We extracted 48 PARGs expression profile and then conducted differentially expressed analysis, following building a Cox model to predict the survival of HNSCC patients. Subsequently, the relationships between the risk score, immune landscape, chemo-, and immune-therapy responses were analyzed, respectively. Moreover, we investigated the prognostic value, and further predicted the pathways influenced by PARGs. Finally, we identified the biological function of crucial PARGs. A total of 18 differentially expressed PARGs were identified in HNSCC, and a Cox model including CASP8, FADD, NLRP1, TNF, and ZBP1 was constructed, which showed that the risk score was associated with the prognosis as well as immune infiltration of HNSCC patients, and the risk score could be regarded as an independent biomarker. Additionally, patients with high-risk score might be an indicator of lymph node metastasis and advanced clinical stage. High-risk scores also contributed to the chemotherapy resistance and immune escape of HNSCC patients. In addition, FADD and ZBP1 played a crucial role in various cancer-related pathways, such as the MAPK, WNT, and MTOR signaling pathways. On the other hand, we suggested that FADD facilitated the progression and 5-fluorouracil (5-FU) resistance of HNSCC cells. A signature based on PANoptosis showed great predictive power for lymph node metastasis and advanced stage, suggesting that the risk score might be an independent prognostic biomarker for HNSCC. Meanwhile, FADD, identified as a prognostic biomarker, may represent an effective therapeutic target for HNSCC.

Bone Quality as Measured by Hounsfield Units More Accurately Predicts Proximal Junctional Kyphosis than Vertebral Bone Quality Following Long-Segment Thoracolumbar Fusion

To compare the prognostic power of Hounsfield units (HU) and Vertebral Bone Quality (VBQ) score for predicting proximal junctional kyphosis (PJK) following long-segment thoracolumbar fusion to the upper thoracic spine (T1-T6). Vertebral bone quality around the upper instrumented vertebrae (UIV) was measured using HU on preoperative CT and VBQ on preoperative MRI. Spinopelvic parameters were also categorized according to the Scoliosis Research Society-Schwab classification. Univariable analysis to identify predictors of the occurrence of PJK and survival analyses with Kaplan-Meier method and Cox regression were performed to identify predictors of time to PJK (defined as ≥10° change in Cobb angle of UIV+2 and UIV). Sensitivity analyses showed thresholds of HU < 164 and VBQ > 2.7 to be most predictive for PJK. Seventy-six patients (mean age 66.0 ± 7.0 years; 27.6% male) were identified, of whom 15 suffered PJK. Significant predictors of PJK were high postoperative pelvic tilt (P= 0.038), high postoperative T1-pelvic angle (P= 0.041), and high postoperative PI-LL mismatch (P= 0.028). On survival analyses, bone quality, as assessed by the average HU of the UIV and UIV+1 was the only significant predictor of time to PJK (odds ratio [OR] 3.053; 95% CI 1.032-9.032; P= 0.044). VBQ measured using the UIV, UIV+1, UIV+2, and UIV-1 vertebrae approached, but did not reach significance (OR 2.913; 95% CI 0.797-10.646; P= 0.106). In larger cohorts, VBQ may prove to be a significant predictor of PJK following long-segment thoracolumbar fusion. However, Hounsfield units on CT have greater predictive power, suggesting preoperative workup for long-segment thoracolumbar fusion benefits from computed tomography versus magnetic resonance imaging alone to identify those at increased risk of PJK.

Identification and validation of a gap junction protein related signature for predicting the prognosis of renal clear cell carcinoma.

Gap junction proteins (GJPs) are a class of channel proteins that are closely related to cell communication and tumor development. The objective of this study was to screen out GJPs related prognostic signatures (GRPS) associated with clear cell renal cell carcinoma (ccRCC). GJPs microarray data for ccRCC patients were obtained from The Gene Expression Omnibus (GEO) database, along with RNA sequencing data for tumor and paired normal tissues from The Cancer Genome Atlas (TCGA) database. In the TCGA database, least absolute shrinkage and selection Operator (LASSO) and Cox regression models were used to identify GJPs with independent prognostic effects as GRPS in ccRCC patients. According to the GRPS expression and regression coefficient from the multivariate Cox regression model, the risk score (RS) of each ccRCC patient was calculated, to construct the RS prognostic model to predict survival. Overall survival (OS) and progression-free survival (PFS) analyses; gene pan-cancer analysis; single gene survival analysis; gene joint effect analysis; functional enrichment analysis; tumor microenvironment (TME) analysis; tumor mutational burden (TMB) analysis; and drug sensitivity analysis were used to explore the biological function, mechanism of action and clinical significance of GRPS in ccRCC. Further verification of the genetic signature was performed with data from the GEO database. Finally, the cytofunctional experiments were used to verify the biological significance of GRPS associated GJPs in ccRCC cell lines. GJA5 and GJB1, which are GRPS markers of ccRCC patients, were identified through LASSO and Cox regression models. Low expression of GJA5 and GJB1 is associated with poor patient prognosis. Patients with high-RS had significantly shorter OS and PFS than patients with low-RS (p< 0.001). The risk of death for individuals with high-RS was 1.695 times greater than that for those with low-RS (HR = 1.695, 95%CI= 1.439-1.996, p< 0.001). Receiver Operating Characteristic (ROC) curve showed the great predictive power of the RS prognostic model for the survival rate of patients. The area under curve (AUC) values for predicting 1-year, 3-year and 5-year survival rates were 0.740, 0.781 and 0.771, respectively. The clinical column chart was also reliable for predicting the survival rate of patients, with AUC values of 0.859, 0.846 and 0.796 for predicting 1-year, 3-year and 5-year survival, respectively. The GRPS was associated with immune cell infiltration, the TME, the TMB, and sensitivity to chemotherapy drugs. Further in vitro experiments showed that knockdown of GJA5 or GJB1 could promote the proliferation, migration and epithelial-mesenchymal transition (EMT) and inhibit apoptosis of ccRCC cells. GJA5 and GJB1 could be potential biological markers for predicting survival in patients with ccRCC.

Single nucleotide polymorphism based on hypertension potential risk prediction using LSTM with Adam optimizer

&lt;div align="center"&gt;&lt;span&gt;Recent healthcare research has focused a great deal of interest on using genetic data analysis to predict the risk of hypertension. This paper presents a unique method for accurately predicting the vulnerability to hypertension by utilizing single nucleotide polymorphism (SNP) data. We present a novel neural network design utilizing the adaptive moment (Adam) optimizer to describe the intricate temporal correlations in SNPs. The study used a dataset with carefully preprocessed SNP data from a broad cohort for model input. The long short-term memory (LSTM) network was methodically built and trained with hyper-parameter and fine-tuning using the Adam optimizer to converge on ideal weights. Our findings indicate encouraging predictive performance, highlighting the suggested methodology’s usefulness in determining hypertension risk factors. The result showed that the proposed method achieved stability in the performance of 89% accuracy, 96% precision, 88% recall, and 92% F1-score. Due to its higher accuracy and greater predictive power, our SNP-based LSTM methodology is superior to the conventional machine learning method. By providing a novel framework that uses genetic data to predict the risk of hypertension, this research makes substantial contribution to the field of predictive healthcare. This framework helps with early intervention and customized preventative efforts.&lt;/span&gt;&lt;/div&gt;

Performance comparison of stress hyperglycemia ratio for predicting fatal outcomes in patients with thrombolyzed acute ischemic stroke.

The stress hyperglycemia ratio (SHR), a newly developed metric, is used to assess adverse outcomes in patients with acute ischemic stroke (AIS). However, the relationship between SHR and fatal outcomes (in-hospital mortality [IHM], malignant cerebral edema [MCE], symptomatic intracerebral hemorrhage [sICH], 3-month mortality, and poor functional outcome) in AIS patients receiving recombinant tissue plasminogen activator (rt-PA) treatment is unclear, and determining the optimal threshold remains incomplete. We retrospectively enrolled a total of 345 AIS patients treated with rt-PA during 2015-2022 and collected data on various glucose metrics, including different types of SHR, glycemic gap (GG), random plasma glucose (RPG), fasting plasma glucose (FPG), and hemoglobin A1c (HbA1c). SHR and GG were calculated using these equations: SHR1, [FPG]/[HbA1c]; SHR2, [admission RPG]/[HbA1c]; SHR3, FPG/[(1.59 × HbA1c)-2.59]; SHR4, [admission RPG]/[(1.59 × HbA1c)-2.59]; GG, admission RPG - [(1.59 × HbA1c)-2.59]. We used multivariable logistic regression analysis (MVLR) to identify the association between different glucose metrics and outcomes while comparing their predictive values. SHR1 had the greatest predictive power and a more significant correlation with fatal outcomes than other continuous glucose metrics. The area under the curve of the SHR1 for IHM, MCE, and sICH, 3-month mortality, and poor functional outcome were 0.75, 0.77, 0.77, 0.76, and 0.73, respectively. SHR1 (per 1-point increases) was independently associated with IHM (Odds ratios [ORs] = 5.80; 95% CI [1.96, 17.17]; p = 0.001), MCE (ORs = 4.73; 95% CI [1.71, 13.04]; p = 0.003), sICH (ORs = 4.68, 95% CI [1.48-14.82]; p = 0.009), 3-month mortality (ORs = 10.87; 95% CI [3.56, 33.21]; p<0.001), and 3-month poor functional outcome (ORs = 8.05; 95% CI [2.77, 23.39]; p<0.001) after adjustment in MVLR. In subgroup analysis, elevated SHR1 was associated with fatal outcomes in patients with non-diabetes, SBP≥ 180 mmHg, and NIHSS <16. SHR1 demonstrates an independent association with fatal outcomes in AIS patients treated with rt-PA, exhibiting superior predictive ability over other glucose metrics.

It’s Not Always about Wide and Deep Models: Click-Through Rate Prediction with a Customer Behavior-Embedding Representation

Alongside natural language processing and computer vision, large learning models have found their way into e-commerce. Especially, for recommender systems and click-through rate prediction, these models have shown great predictive power. In this work, we aim to predict the probability that a customer will click on a given recommendation, given only its current session. Therefore, we propose a two-stage approach consisting of a customer behavior-embedding representation and a recurrent neural network. In the first stage, we train a self-supervised skip-gram embedding on customer activity data. The resulting embedding representation is used in the second stage to encode the customer sequences which are then used as input to the learning model. Our proposed approach diverges from the prevailing trend of utilizing extensive end-to-end models for click-through rate prediction. The experiments, which incorporate a real-world industrial use case and a widely used as well as openly available benchmark dataset, demonstrate that our approach outperforms the current state-of-the-art models. Our approach predicts customers’ click intention with an average F1 accuracy of 94% for the industrial use case which is one percentage point higher than the state-of-the-art baseline and an average F1 accuracy of 79% for the benchmark dataset, which outperforms the best tested state-of-the-art baseline by more than seven percentage points. The results show that, contrary to current trends in that field, large end-to-end models are not always needed. The analysis of our experiments suggests that the reason for the performance of our approach is the self-supervised pre-trained embedding of customer behavior that we use as the customer representation.

Prediction of patient admission and readmission in adults from a Colombian cohort with bipolar disorder using artificial intelligence.

Bipolar disorder (BD) is a chronically progressive mental condition, associated with a reduced quality of life and greater disability. Patient admissions are preventable events with a considerable impact on global functioning and social adjustment. While machine learning (ML) approaches have proven prediction ability in other diseases, little is known about their utility to predict patient admissions in this pathology. To develop prediction models for hospital admission/readmission within 5 years of diagnosis in patients with BD using ML techniques. The study utilized data from patients diagnosed with BD in a major healthcare organization in Colombia. Candidate predictors were selected from Electronic Health Records (EHRs) and included sociodemographic and clinical variables. ML algorithms, including Decision Trees, Random Forests, Logistic Regressions, and Support Vector Machines, were used to predict patient admission or readmission. Survival models, including a penalized Cox Model and Random Survival Forest, were used to predict time to admission and first readmission. Model performance was evaluated using accuracy, precision, recall, F1 score, area under the receiver operating characteristic curve (AUC) and concordance index. The admission dataset included 2,726 BD patients, with 354 admissions, while the readmission dataset included 352 patients, with almost half being readmitted. The best-performing model for predicting admission was the Random Forest, with an accuracy score of 0.951 and an AUC of 0.98. The variables with the greatest predictive power in the Recursive Feature Elimination (RFE) importance analysis were the number of psychiatric emergency visits, the number of outpatient follow-up appointments and age. Survival models showed similar results, with the Random Survival Forest performing best, achieving an AUC of 0.95. However, the prediction models for patient readmission had poorer performance, with the Random Forest model being again the best performer but with an AUC below 0.70. ML models, particularly the Random Forest model, outperformed traditional statistical techniques for admission prediction. However, readmission prediction models had poorer performance. This study demonstrates the potential of ML techniques in improving prediction accuracy for BD patient admissions.