Cross-validation Method Research Articles

Method of generating potential evapotranspiration with high precision and resolution

Potential evapotranspiration (PET) is a key factor in hydrological cycle and energy balance and plays an important role in drought and global climate change response. Existing observational and modeling methods for PET retrieval have their limitations, such as low precision and poor spatial resolution, which becomes the focus of this study. A hybrid PET fusion (HPF) method is proposed by fusing station- and grid-based PET, in which the PET expression is determined by considering the factors of location, temperature, and zenith total delay (ZTD). In addition, an improved Helmert variance component estimation method is introduced to determine the optimal weights of the HPF model. Corresponding data, which include monthly Thornthwaite (TH)-derived PET data with a spatial resolution of 0.25° × 0.25° and Penman–Monteith (PM)-derived PET data at 704 meteorological stations, over the past 60 years from 1959 to 2018 in China are selected. The 10-fold cross-validation method is introduced to evaluate the internal and external accuracies of the proposed HPF method. Statistical result shows that the average root mean square (RMS) of the proposed HPF method is 13.98 mm, with an average RMS improvement rate (IR) of 46.71 % compared with TH-derived PET, when PM-derived PET is regarded as a reference. Moreover, the performance of the HPF-derived standardized precipitation evapotranspiration index (SPEI) is evaluated at different time scales, and the average RMS is 0.3, with an average RMS IR of 26.33 % compared with TH-derived SPEI. Such results verify the good performance of the proposed HPF model and enrich the methods for obtaining PET with high precision and resolution.

DDLA: a double deep latent autoencoder for diabetic retinopathy diagnose based on continuous glucose sensors.

The current diagnosis of diabetic retinopathy is based on fundus images and clinical experience. However, considering the ineffectiveness and non-portability of medical devices, we aimed to develop a diagnostic model for diabetic retinopathy based on glucose series data from the wearable continuous glucose monitoring system. Therefore, this study developed a novel method, i.e., double deep latent autoencoder, for exploring glycemic variability influence from multi-day glucose data for diabetic retinopathy. Specifically, the model proposed in this research could encode continuous glucose sensor data with non-continuous and variable length via the integration of a data reorganization module and a novel encoding module with fragmented-missing-wise objective function. Additionally, the model implements a double deep autoencoder, which integrated convolutional neural network, long short-term memory, to jointly capturing the inter-day and intra-day glucose latent features from glucose series. The effectiveness of the proposed model is evaluated through a cross-validation method to clinical datasets of 765 type 2 diabetes patients. The proposed method achieves the highest accuracy value (0.89), precision value (0.88), and F1 score (0.73). The results suggest that our model can be used to remotely diagnose and screen for diabetic retinopathy by learning potential features of glucose series data collected by wearable continuous glucose monitoring systems.

METHODS OF FORECASTING GRAIN CROP YIELD INDICATORS TAKING INTO ACCOUNT THE INFLUENCE OF METEOROLOGICAL CONDITIONS IN THE INFORMATION-ANALYTICAL SUBSYSTEM

Forecasting crop yields is one of the key challenges for the agricultural sector, especially in the context of a changing climate and unstable weather conditions. Kazakhstan, possessing significant territories suitable for growing grain crops, faces many challenges related to the effective management of agricultural activities. In this regard, yield forecasting becomes an integral part of planning and decision-making processes in agriculture. Information and analytical subsystems that integrate yield forecasting methods allow agribusinesses to estimate future production more accurately, minimise risks associated with climate change and optimise resource use. An important component of such systems is the consideration of weather conditions, as weather factors have a direct impact on crop growth and development. The purpose of this article is to develop and evaluate modern methods of forecasting grain yields taking into account the influence of weather conditions, as well as their integration into information-analytical subsystems to improve the accuracy of agricultural forecasting. To achieve this goal, the article addresses the following tasks: to analyse existing methods of yield forecasting and identify their advantages and disadvantages, to develop forecasting models, including machine learning methods such as gradient bousting and recurrent neural networks, to validate the developed models on the basis of historical data using cross-validation methods, to evaluate the effectiveness of the proposed methods and compare them with basic models such as linear regression and simple average, to evaluate the effectiveness of the proposed methods and to compare them with the basic models such as linear regression and simple average. This article reviews modern methods of forecasting grain crop yields in Kazakhstan, as well as technologies used in information-analytical subsystems. Particular attention is paid to the analysis of the influence of meteorological conditions on yields and the development of models that take this factor into account. The presented review and research results are aimed at improving the existing approaches to the management of agricultural processes under conditions of growing uncertainty caused by climate change. The article explores an important scientific task related to the development of methods for step-by-step forecasting of agrometeorological factors and grain yields, relying on the principle of analogy.

Comparison of Classification Results of SVM, KNN, Decision Tree, and Ensemble Methods in Diabetes Diagnosis

This study aims to determine which algorithms and test techniques are the most optimal in detecting diabetes mellitus and obtaining the best results based on the value of accuracy, precision, and recall. In this study, approaches were used in early diagnosis of diabetes using KNN, SVM, Decision Tree, and Ensemble Majority Voting methods in Percentage Split and K-Fold Cross Validation methods. Diabetes is a disease characterized by high blood sugar (glucose) levels and can cause a variety of disease complications and damage to the body's organs if not treated immediately. Early diagnosis of diabetes is becoming crucial so that people can take immediate action to the hospital for immediate treatment. The data used is Healthcare-Diabetes from Kaggle. The results of this study have found that the K-Fold Cross Validation method is better because it can provide an average improvement in Ensemble accuracy of 13.42% compared to the Percentage Split method which only gives an average increase in Ensamble accuracy of 9.15%. The best algorithm for classifying diabetes disease is the Ensemble Majority Voting algorithm using the K-Fold Cross Validation method with a 98.81% accuracy rate. These excellent research results may contribute to detecting early symptoms of diabetes before it become too severe.

Variable selection for uncertain regression models based on elastic net method

On one hand, determining the exact number of explanatory variables in uncertain regression analysis is often challenging, as real-world scenarios are interconnected. On the other hand, it’s worth noting that even when the number of variables is determined, some variables may have minimal impact on the response variable, and retaining them would only increase the complexity of the model. Therefore, variable selection methods can help eliminate variables with minimal impact to enhance the simplicity and interpretability of the model. In light of this, this paper first introduces a novel variable selection strategy based on the elastic net method to address the variable selection problem in uncertain regression models. This method not only offers the flexibility to simplify uncertain regression analysis into Lasso estimation or Ridge estimation but also exhibits extremely high efficiency in variable selection. Secondly, to address the issue of determining tuning parameters, we introduce the cross-validation method to ensure optimal model performance. Furthermore, we propose an uncertain hypothesis testing method based on imprecise observational data, which can help us assess the suitability of estimated parameters. Finally, through a series of numerical examples, we validate the feasibility and effectiveness of this method.

Vascular Biomarkers for Pulmonary Nodule Malignancy: Arteries vs. Veins.

This study aims to investigate the association between the arteries and veins surrounding a pulmonary nodule and its malignancy. A dataset of 146 subjects from a LDCT lung cancer screening program was used in this study. AI algorithms were used to automatically segment and quantify nodules and their surrounding macro-vasculature. The macro-vasculature was differentiated into arteries and veins. Vessel branch count, volume, and tortuosity were quantified for arteries and veins at different distances from the nodule surface. Univariate and multivariate logistic regression (LR) analyses were performed, with a special emphasis on the nodules with diameters ranging from 8 to 20 mm. ROC-AUC was used to assess the performance based on the k-fold cross-validation method. Average feature importance was evaluated in several machine learning models. The LR models using macro-vasculature features achieved an AUC of 0.78 (95% CI: 0.71-0.86) for all nodules and an AUC of 0.67 (95% CI: 0.54-0.80) for nodules between 8-20 mm. Models including macro-vasculature features, demographics, and CT-derived nodule features yielded an AUC of 0.91 (95% CI: 0.87-0.96) for all nodules and an AUC of 0.82 (95% CI: 0.71-0.92) for nodules between 8-20 mm. In terms of feature importance, arteries within 5.0 mm from the nodule surface were the highest-ranked among macro-vasculature features and retained their significance even with the inclusion of demographics and CT-derived nodule features. Arteries within 5.0 mm from the nodule surface emerged as a potential biomarker for effectively discriminating between malignant and benign nodules.

The early prognostic value of the 1–4-day BCM/PA trend after admission in neurocritical patients

The purpose of this study was to investigate early stage dynamic changes in relevant indicators in neurocritical patients to identify biomarkers that can predict a poor prognosis at an early stage (1–4 days after admission). This study retrospectively collected clinical data, inflammatory indicators, and nutritional indicators from 77 patients at the neurology intensive care unit. The 3-month modified Rankin scale score was used as the outcome indicator. A linear mixed model was used to analyze changes in inflammatory indicators and nutritional indicators in neurocritical patients over time from 1–4 days after admission. Logistic regression was used to determine the independent risk factors for a poor prognosis in neurocritical patients and to construct a predictive model. The predictive efficacy of the model was verified using leave-one-out cross-validation and decision curve analysis methods. The analysis results showed that 1–4 days after admission, the inflammatory indicators of white blood cell and absolute monocyte counts and the nutritional indicators of body cell mass(BCM), fat-free mass, body cell mass/phase angle (BCM/PA), intracellular water, extracellular water, and skeletal muscle index increased overall, while the nutritional indicators of albumin and visceral fat area decreased overall. The logistic multivariate regression model showed that the Charlson comorbidity index (CCI) (odds ratio (OR) = 2.526, 95% CI [1.202, 5.308]), hemoglobin (Hb)(on admission)-Hb(min) (OR = 1.049, 95% CI [1.015, 1.083), BCM(on admission) (OR = 0.794, 95% CI [0.662, 0.952]), and the change in BCM/PA 1–4 days after admission (OR = 1.157, 95% CI [1.070, 1.252]) were independent risk factors for a poor prognosis in neurocritical patients. The predictive analysis showed that the predictive power of Model 1 with BCM/PA (area under the curve (AUC) = 0.95, 95% CI (0.90, 0.99)) was 93%, 65%, 141%, and 133% higher than that of Model 2 without BCM/PA, the CCI, the APACHE II score, and the NRS2002 score (all P < 0.05), respectively. The CCI, Hb(on admission)-Hb(min), BCM(on admission), and an increase in BCM/PA 1–4 days after admission were independently associated with a poor prognosis in neurocritical patients. Of these variables, BCM/PA may be a valid indicator for early stage prediction of a poor prognosis in neurocritical patients.

Deep learning-assisted distinguishing breast phyllodes tumours from fibroadenomas based on ultrasound images: a diagnostic study.

To evaluate the performance of ultrasound-based deep learning (DL) models in distinguishing breast phyllodes tumours (PTs) from fibroadenomas (FAs) and their clinical utility in assisting radiologists with varying diagnostic experiences. We retrospectively collected 1180 ultrasound images from 539 patients (247 PTs and 292 FAs). Five DL network models with different structures were trained and validated using nodule regions annotated by radiologists on breast ultrasound images. DL models were trained using the methods of transfer learning and 3-fold cross-validation. The model demonstrated the best evaluation index in the 3-fold cross-validation was selected for comparison with radiologists' diagnostic decisions. Two-round reader studies were conducted to investigate the value of DL model in assisting 6 radiologists with different levels of experience. Upon testing, Xception model demonstrated the best diagnostic performance (area under the receiver-operating characteristic curve: 0.87; 95% CI, 0.81-0.92), outperforming all radiologists (all P < .05). Additionally, the DL model enhanced the diagnostic performance of radiologists. Accuracy demonstrated improvements of 4%, 4%, and 3% for senior, intermediate, and junior radiologists, respectively. The DL models showed superior predictive abilities compared to experienced radiologists in distinguishing breast PTs from FAs. Utilizing the model led to improved efficiency and diagnostic performance for radiologists with different levels of experience (6-25 years of work). We developed and validated a DL model based on the largest available dataset to assist in diagnosing PTs. This model has the potential to allow radiologists to discriminate 2 types of breast tumours which are challenging to identify with precision and accuracy, and subsequently to make more informed decisions about surgical plans.

Accurate Machine Learning-based Monitoring of Anesthesia Depth with EEG Recording.

General anesthesia, pivotal for surgical procedures, requires precise depth monitoring to mitigate risks ranging from intraoperative awareness to postoperative cognitive impairments. Traditional assessment methods, relying on physiological indicators or behavioral responses, fall short of accurately capturing the nuanced states of unconsciousness. This study introduces a machine learning-based approach to decode anesthesia depth, leveraging EEG data across different anesthesia states induced by propofol and esketamine in rats. Our findings demonstrate the model's robust predictive accuracy, underscored by a novel intra-subject dataset partitioning and a 5-fold cross-validation method. The research diverges from conventional monitoring by utilizing anesthetic infusion rates as objective indicators of anesthesia states, highlighting distinct EEG patterns and enhancing prediction accuracy. Moreover, the model's ability to generalize across individuals suggests its potential for broad clinical application, distinguishing between anesthetic agents and their depths. Despite relying on rat EEG data, which poses questions about real-world applicability, our approach marks a significant advance in anesthesia monitoring.

Multimodal registration network with multi-scale feature-crossing.

A critical piece of information for prostate intervention and cancer treatment is provided by the complementary medical imaging modalities of ultrasound (US) and magnetic resonance imaging (MRI). Therefore, MRI-US image fusion is often required during prostate examination to provide contrast-enhanced TRUS, in which image registration is a key step in multimodal image fusion. We propose a novel multi-scale feature-crossing network for the prostate MRI-US image registration task. We designed a feature-crossing module to enhance information flow in the hidden layer by integrating intermediate features between adjacent scales. Additionally, an attention block utilizing three-dimensional convolution interacts information between channels, improving the correlation between different modal features. We used 100 cases randomly selected from The Cancer Imaging Archive (TCIA) for our experiments. A fivefold cross-validation method was applied, dividing the dataset into five subsets. Four subsets were used for training, and one for testing, repeating this process five times to ensure each subset served as the test set once. We test and evaluate our technique using fivefold cross-validation. The cross-validation trials result in a median target registration error of 2.20 mm on landmark centroids and a median Dice of 0.87 on prostate glands, both of which were better than the baseline model. In addition, the standard deviation of the dice similarity coefficient is 0.06, which suggests that the model is stable. We propose a novel multi-scale feature-crossing network for the prostate MRI-US image registration task. A random selection of 100 cases from The Cancer Imaging Archive (TCIA) was used to test and evaluate our approach using fivefold cross-validation. The experimental results showed that our method improves the registration accuracy. After registration, MRI and TURS images were more similar in structure and morphology, and the location and morphology of cancer were more accurately reflected in the images.

PSXI-23 Development of sub-models to estimate protein requirements and supply of lactating dairy cows using machine learning algorithms

Abstract The objectives of this study were to develop sub-models for predicting protein requirements and supply, encompassing 1) net protein for maintenance (NPm), 2) lactation (NPl), 3) rumen undegradable protein (RUP), and 4) duodenal microbial nitrogen (MicN) from the feed protein. The dataset used in this study was constructed by integrating in vivo experimental data collected from open databases (the National Animal Nutrition Program) and articles (Journal of Dairy Science), which includes a total of 1,779 observations from 436 publications. In the development of the model, animal information and feed chemical components were used as candidate variables, and two types of machine learning algorithms, Random Forest Regression (RFR) and Support Vector Regression (SVR) were employed. After testing, the following predictors were selected for predicting: 1) NPm: body weight (BW) and dry matter intake (DMI), 2) NPl: BW, DMI, days in milk (DIM), and dietary organic matter (OM) and crude protein (CP) contents, 3) RUP: DIM, DMI, dietary DM content, and CP fraction intake (B and C), and 4) MicN: DIM, DMI, DM, dietary neutral detergent fiber (NDF) content, CP fraction intake (A, B, and C). The selected models were assessed using a cross-validation method with the following statistical metrics including the coefficient of determination (R2), root-mean-square error of prediction (RMSEP), residual analyses, and concordance correlation coefficient (CCC). For the RUP and MicN models, they were compared with the NASEM (2021) model. In predicting NPm, both SVR and RFR algorithms demonstrated increased precision (R2 = 0.965 vs. 0.969) and accuracy (RMSEP = 9.7 vs. 9.2 g/d); however, during residual analysis, the RFR model showed a statistically significant slope bias (P &amp;lt; 0.05). In the NPl prediction, the RFR algorithm showed slightly greater performance compared with SVR (R2 = 0.864 vs. 0.814 and RMSEP = 86.7 vs. 98.5 g/d). However, similar to the NPm prediction, the RFR model displayed a statistically significant slope bias (P &amp;lt; 0.05). In the supply model, the RFR model exhibited the greatest precision and accuracy in predicting RUP (R2 = 0.60, RMSEP = 0.326 kg/d, and CCC = 0.71) without any biases. This model achieved a 2.37-fold increase in R2 and a decrease of 0.111 kg/d in RMSEP compared with the NASEM model. As for MicN prediction, the SVR model performed the best (R2 = 0.76, RMSEP = 42.4 g/d, and CCC = 0.86) without biases. This model attained a 19-times improvement in R2 and a reduction of 38.7 g/d in RMSEP when compared with the NASEM model. In conclusion, the models developed using machine learning algorithms can be helpful for accurately and precisely predicting protein requirements and supply based on animal information and the chemical composition of feed.

A deep learning method for contactless emotion recognition from ballistocardiogram

Emotion recognition is a major research point in the field of affective computing. Existing research on the application of physiological signals to emotion recognition mainly focuses on the processing of contact signals. However, there are issues with contact signal acquisition equipment, such as limited portability and poor user compliance, which make it difficult to promote its use. To explore a new method for emotion recognition based on contactless ballistocardiogram (BCG), we proposed a SE-CNN model with a multi-class focal loss function. To construct the dataset, we used audio-visual stimuli to evoke the subjects’ emotions and collected data on the subjects’ three discrete emotions, positive, neutral, and negative, through our established BCG signal acquisition system based on a piezoelectric ceramics sensor. Root mean square filter and thresholding were used to detect and eliminate motion artifacts of BCG signals. We did two kinds of preprocessing on BCG signals: wavelet transform and bandpass filtering, to explore the effect of different components of BCG on emotion recognition. Subsequently, we verified the model’s performance and cross-time working ability through traditional K-Fold and our proposed K-Session cross-validation methods. The results showed that the band-pass filtering method was more beneficial to the current classification task. Under K-Fold cross-validation, the model’s accuracy, precision, and recall were 97.21%, 97.00%, and 97.11%. Under K-Session cross-validation, the model’s accuracy, precision, and recall were 94.66%, 93.92%, and 94.86%, respectively, all of which were better than the classification effect of synchronous ECG. The reliability of BCG in contactless emotion recognition was proved.

Making the most out of timeseries symptom data: A machine learning study on symptom predictions of internet-based CBT

ObjectivePredicting who will not benefit enough from Internet-Based Cognitive Behavioral (ICBT) Therapy early on can assist in better allocation of limited mental health care resources. Repeated measures of symptoms during treatment is the strongest predictor of outcome, and we want to investigate if methods that explicitly account for time-dependency are superior to methods that do not, with data from (a) only two pre-treatment timepoints and (b) the pre-treatment timepoints and three timepoints during initial treatment. MethodsWe use 1) commonly used time-independent methods (i.e., Linear Regression and Random Forest models) and 2) time-dependent methods (i.e., multilevel model regression, mixed-effects random forest, and a Long Short-Term Memory model) to predict symptoms during treatment, including the final outcome. This is done with symptom scores from 6436 ICBT patients from regular care, using robust multiple imputation and nested cross-validation methods. ResultsThe models had a 14 %–12 % root mean squared error (RMSE) in predicting the post-treatment outcome, corresponding to a balanced accuracy of 67–74 %. Time-dependent models did not have higher accuracies. Using data for the initial treatment period (b) instead of only from before treatment (a) increased prediction results by 1.3 % percentage points (12 % to 10.7 %) RMSE and 6 % percentage points BACC (69 % to 75 %). ConclusionTraining prediction models on only symptom scores of the first few weeks is a promising avenue for symptom predictions in treatment, regardless of which model is used. Further research is necessary to better understand the interaction between model complexity, dataset length and width, and the prediction tasks at hand.

Prediksi Harga Crypto dengan Algoritma Jaringan Saraf Tiruan

Cryptocurrency has emerged as a crucial element in the global financial market, prompting investor interest in predicting prices for informed investment decision-making. However, the highly volatile nature of cryptocurrency prices has made trading in the crypto market speculative and laden with significant risks. Therefore, this study aims to apply Artificial Neural Networks (ANNs) using the Long Short-Term Memory (LSTM) algorithm to forecast cryptocurrency prices. This method leverages the ANN's ability to recognize complex patterns and trends in historical cryptocurrency price data. The research findings demonstrate significant accuracy levels for three types of crypto: BTC (86.86%), BNB (96.8%), and Doge (97%). Evaluation was conducted using the k-fold cross-validation method with k=5, where data was divided into five equally sized groups. Accuracy was computed by comparing actual prices with those predicted by the LSTM model. This evaluation approach provides robust insights into the LSTM model's effectiveness in predicting cryptocurrency prices, considering significant data variations. The predicted values from this study are BTC at $4306, BNB at $58.7, and Dogecoin at $0.037. These implications underscore the critical role of the number of epochs in influencing the performance and accuracy of cryptocurrency price predictions. Thus, this research is poised to offer deeper insights into cryptocurrency market behavior and provide practical guidance for investors to make more informed and measured investment decisions.

Construction of a clinical prediction model for the diagnosis of immune thrombocytopenia based on clinical laboratory parameters

PurposePrimary immune thrombocytopenia (ITP) is an autoimmune bleeding disorder characterized by isolated thrombocytopenia that is often misdiagnosed due to the lack of a gold standard for diagnosis and currently relies on exclusionary approaches. This project combines several laboratory parameters to construct a clinical prediction model for adult ITP patients.MethodsA total of 428 patients with thrombocytopenia who visited the West China Hospital of Sichuan University between January 2021 and March 2023 were enrolled. Based on the diagnostic criteria, we divided those patients into an ITP group and a non-ITP group. A total of 34 laboratory parameters were analyzed via univariate analysis and correlation analysis, and the least absolute shrinkage and selection operator regression analysis was used to establish the model. The training and validation sets were divided at a ratio of 7:3, and we used a fivefold cross-validation method to construct the model.ResultsThe model included the following variables: red blood cell, mean corpuscular hemoglobin concentration, red blood cell distribution width-standard deviation, platelet variability index score, immature platelet fraction, lymphocyte absolute value. The prediction model exhibited good performance, with a sensitivity of 0.89 and a specificity of 0.83 in the training set and a sensitivity of 0.90 and a specificity of 0.87 in the validation set.ConclusionThe clinical prediction model can assess the probability of ITP in thrombocytopenic patients and has good predictive accuracy for the diagnosis of ITP.

A Machine-Learning-Based Method for Identifying the Failure Risk State of Fissured Sandstone under Water-Rock Interaction.

The mechanical properties of fissured sandstone will deteriorate under water-rock interaction. It is crucial to extract the precursor information of fissured sandstone instability under water-rock interaction. The potential of each acoustic emission (AE) parameter as a precursor for instability in the failure process of fissured sandstone was investigated in this study. An experimental dataset comprising 586 acoustic emission experiments was established, and subsequent classification training and testing were conducted using three machine learning (ML) models: AdaBoost, MLP, and Random Forest (RF). The primary parameters for identifying the instability risk state of fissured sandstone include acoustic emission ringing count, energy (mV·ms), centroid frequency, peak frequency, Rise Angle (RA), Average Frequency (AF), b value, and the natural/saturated state of fissured sandstone: state. To enhance data utilization, a 10-fold cross-validation method was employed during the model training process. The machine learning models were developed and designed to identify the instability risk of fissured sandstone under the natural and saturated states. The results demonstrated that the established RF model was capable of identifying fissured sandstone instability risks with an accuracy of 97.87%. Feature importance analysis revealed that state and b value exerted the most significant influence on identification results. The Spearman correlation coefficient was utilized to assess the correlation between input features. This study can provide technical support to identify the risk of instability of fissured sandstones under both natural and saturated water conditions. Based on the models developed in this study, it is possible to implement an early warning method for instability in fissured sandstone that meets realistic working conditions. Compared with the traditional empirical and formulaic methods, the machine learning method can more quickly process huge amounts of AE data and accurately identify the damage state of fissured sandstone.

Predicting splicing patterns from the transcription factor binding sites in the promoter with deep learning

BackgroundAlternative splicing is a pivotal mechanism of post-transcriptional modification that contributes to the transcriptome plasticity and proteome diversity in metazoan cells. Although many splicing regulations around the exon/intron regions are known, the relationship between promoter-bound transcription factors and the downstream alternative splicing largely remains unexplored.ResultsIn this study, we present computational approaches to unravel the regulatory relationship between promoter-bound transcription factor binding sites (TFBSs) and the splicing patterns. We curated a fine dataset that includes DNase I hypersensitive site sequencing and transcriptomes across fifteen human tissues from ENCODE. Specifically, we proposed different representations of TF binding context and splicing patterns to examine the associations between the promoter and downstream splicing events. While machine learning models demonstrated potential in predicting splicing patterns based on TFBS occupancies, the limitations in the generalization of predicting the splicing forms of singleton genes across diverse tissues was observed with carefully examination using different cross-validation methods. We further investigated the association between alterations in individual TFBS at promoters and shifts in exon splicing efficiency. Our results demonstrate that the convolutional neural network (CNN) models, trained on TF binding changes in the promoters, can predict the changes in splicing patterns. Furthermore, a systemic in silico substitutions analysis on the CNN models highlighted several potential splicing regulators. Notably, using empirical validation using K562 CTCFL shRNA knock-down data, we showed the significant role of CTCFL in splicing regulation.ConclusionIn conclusion, our finding highlights the potential role of promoter-bound TFBSs in influencing the regulation of downstream splicing patterns and provides insights for discovering alternative splicing regulations.

Bayesian-optimized extreme gradient boosting models for classification problems: an experimental analysis of product return case

PurposeProduct returns are a major challenge for e-businesses as they involve huge logistical and operational costs. Therefore, it becomes crucial to predict returns in advance. This study aims to evaluate different genres of classifiers for product return chance prediction, and further optimizes the best performing model.Design/methodology/approachAn e-commerce data set having categorical type attributes has been used for this study. Feature selection based on chi-square provides a selective features-set which is used as inputs for model building. Predictive models are attempted using individual classifiers, ensemble models and deep neural networks. For performance evaluation, 75:25 train/test split and 10-fold cross-validation strategies are used. To improve the predictability of the best performing classifier, hyperparameter tuning is performed using different optimization methods such as, random search, grid search, Bayesian approach and evolutionary models (genetic algorithm, differential evolution and particle swarm optimization).FindingsA comparison of F1-scores revealed that the Bayesian approach outperformed all other optimization approaches in terms of accuracy. The predictability of the Bayesian-optimized model is further compared with that of other classifiers using experimental analysis. The Bayesian-optimized XGBoost model possessed superior performance, with accuracies of 77.80% and 70.35% for holdout and 10-fold cross-validation methods, respectively.Research limitations/implicationsGiven the anonymized data, the effects of individual attributes on outcomes could not be investigated in detail. The Bayesian-optimized predictive model may be used in decision support systems, enabling real-time prediction of returns and the implementation of preventive measures.Originality/valueThere are very few reported studies on predicting the chance of order return in e-businesses. To the best of the authors’ knowledge, this study is the first to compare different optimization methods and classifiers, demonstrating the superiority of the Bayesian-optimized XGBoost classification model for returns prediction.

Predicting axillary lymph node metastasis in breast cancer based on ultrasound radiofrequency time-series analysis.

The status of axillary lymph nodes (ALN) plays a critical role in the management of patients with breast cancer. It is an urgent demand to develop highly accurate, non-invasive methods for predicting ALN status. To evaluate the efficacy of ultrasound radiofrequency (URF) time-series parameters, in combination with clinical data, in predicting ALN metastasis in patients with breast cancer. We prospectively gathered clinicopathologic and ultrasonic data from patients diagnosed with breast cancer. Various machine-learning (ML) models were developed using all available features to determine the most efficient diagnostic model. Subsequently, distinct prediction models were created using the optimal ML model, and their diagnostic performances were evaluated and compared. The study encompassed 240 patients, of whom 88 had lymph node metastases. A leave-one-out cross-validation (LOOCV) method was used to split the entire dataset into training and testing subsets. The random forest ML model outperformed the other algorithms, with an area under the curve (AUC) of 0.92. Prediction models based on clinical, ultrasonic, URF parameters, clinical + ultrasonic, clinical + URF, and ultrasonic + URF parameters had AUCs of 0.56, 0.79, 0.78, 0.90, 0.80, and 0.84, respectively, in the testing set. The comprehensive diagnostic model (clinical + ultrasonic + URF parameters) demonstrated strong diagnostic capability, with an AUC of 0.94 in the testing set, exceeding any single prediction model. The combined model (clinical + ultrasonic + URF parameters) could be used preoperatively to predict lymph node status, offering valuable input for the design of individualized surgical approaches.

Predictive models, as an idea, to advance the secondary to tertiary transition in science courses

Investigating the transition between the secondary and the tertiary levels is a main theme in mathematics and science education. More so, this paper considers the transition that intersects with the after-effects of COVID-19, or the transition together with an educational context dominated by sociocultural differences and educational disadvantages. With this knowledge in mind, we investigated the effects of predictive mathematical models (multiple regression, logistic regression, and decision trees) to predict &lt;i&gt;at-risk&lt;/i&gt; students at three time intervals (weeks one, three, and seven) in the semester. The idea was implemented with a first-year life science class of 130 students. Variables from an academic readiness questionnaire along with early assessment grades were used to build these models. Through a Monte Carlo cross validation method, the performance of the executed predictive models was assessed, and limitations were reported. We argue that the results obtained from predictive models can support both lecturers and students in the transition phase. The idea can be expanded to other courses in STEM fields and other educational contexts.