Performance Of Machine Learning Models Research Articles

Impact of Feature Selection Techniques on the Performance of Machine Learning Models for Depression Detection Using EEG Data

Major depressive disorder (MDD) poses a significant challenge in mental healthcare due to difficulties in accurate diagnosis and timely identification. This study explores the potential of machine learning models trained on EEG-based features for depression detection. Six models and six feature selection techniques were compared, highlighting the crucial role of feature selection in enhancing classifier performance. This study investigates the six feature selection methods: Elastic Net, Mutual Information (MI), Chi-Square, Forward Feature Selection with Stochastic Gradient Descent (FFS-SGD), Support Vector Machine-based Recursive Feature Elimination (SVM-RFE), and Minimal-Redundancy-Maximal-Relevance (mRMR). These methods were combined with six diverse classifiers: Logistic Regression, Support Vector Machine (SVM), Random Forest, Extreme Gradient Boosting (XGBoost), Categorical Boosting (CatBoost), and Light Gradient Boosting Machine (LightGBM). The results demonstrate the substantial impact of feature selection on model performance. SVM-RFE with SVM achieved the highest accuracy (93.54%) and F1 score (95.29%), followed by Logistic Regression with an accuracy of 92.86% and F1 score of 94.84%. Elastic Net also delivered strong results, with SVM and Logistic Regression both achieving 90.47% accuracy. Other feature selection methods yielded lower performance, emphasizing the importance of selecting appropriate feature selection and machine learning algorithms. These findings suggest that careful selection and application of feature selection techniques can significantly enhance the accuracy of EEG-based depression detection.

HPB SO72 - Predicting Major Intra-operative Blood Loss in Patients undergoing Liver Resection: A Machine Learning Radiomics Analysis Using Pre-Operative Liver CTs

Abstract Background Liver resection offers the best opportunity for curing patients with colorectal metastases and is an important treatment option for patients with hepatocellular carcinoma. However, it can be associated with significant morbidity and mortality, with one of the primary reasons being due to intra-operative blood loss which serves as an important predictor of post-operative outcomes. Predicting blood loss pre-operatively would be a powerful tool for clinicians in identifying higher risk patients. In this study, we examine whether radiomic features of the liver can predict major blood loss (>500 mls) alongside clinical and morphometric features in patients undergoing resection for colorectal metastases. Method Patient data was extracted from a prospectively maintained database of individuals undergoing resection for colorectal liver metastases (CRLMs). Radiomic features were extracted from pre-operative portal venous phase CTs. The background liver was randomly segmented in a two-dimensional plane away from cancer areas. Highly correlated features were dropped, and further feature selection techniques applied. Nested cross validation was performed to train and evaluate the machine learning models with the selected radiomic, clinical, and morphometric features. Results Radiomic features (159) were extracted from 106 patients, and 10 radiomic features comprised of first order and texture features, 3 morphometric, and 4 clinical features were selected for the machine learning models. The models demonstrated a robust ability to discriminate between major and minor blood loss. The best performing machine learning model demonstrated an AUC of 0.76 (95% CI 0.7-0.82). Other performance metrics for this model included a mean accuracy of 0.78 and an F1 Score of 0.87 on the validation data. Conclusion Pre-operative CT radiomic, clinical, and morphometric features have shown a modest ability to predict major intra-operative blood loss in patients undergoing hepatic resection for colorectal liver metastases. Further work is needed to validate and refine these models to improve performance. Additionally, we plan to incorporate data from other centres to validate our findings in further experiments to create a robust tool to guide surgical practice in this patient cohort.

Performance Analysis of ML and DL Models: Impact of Linear and Non-Linear Optimizers on Model Efficiency

Introduction: Emphasizing the effect of linear and non-linear optimizers on the performance of machine learning (ML) and deep learning (DL) models, this work addresses the CIFAR-10 image classification task. The work explores over multiple data sets how optimizers influence model behavior and efficiency. Objectives: With a view on five primary assessment metrics—accuracy, precision, recall, F1-score, and AUC the main purpose is to evaluate on SVM and CNN models the performance of linear optimizers (gradient descent, SGD) and non-linear optimizers (Adam, RMSProp). Methods: The study offers fair and complete comparison by way of a consistent evaluation strategy. Analyzing how different optimizer influences model convergence rates, computational cost, and training stability takes front stage in the experimental setting. Designed on CIFAR-10, optimizers are used on ML and DL models whose performance on all metrics is noted. Results: Results show that non-linear optimizers especially Adam much increase CNN model performance by means of faster convergence, higher classification accuracy, and better model stability. While helpful for simpler models, linear optimizers such SGD show slower convergence and limited adaptation with more complicated data. Conclusions: This study provides direction on selecting appropriate optimizers depending on model complexity, job needs, and computing restrictions coupled with important information on optimizing model training in image classification and other areas.

Machine learning and deep learning models for preoperative detection of lymph node metastasis in colorectal cancer: a systematic review and meta-analysis.

To evaluate the diagnostic performance of Machine Learning (ML) and Deep Learning (DL) models for predicting preoperative Lymph Node Metastasis (LNM) in Colorectal Cancer (CRC) patients. A systematic review and meta-analysis were conducted following PRISMA-DTA and AMSTAR-2 guidelines. We searched PubMed, Web of Science, Embase, and Cochrane Library databases until February 16, 2024. Study quality and risk of bias were assessed using the QUADAS-2 tool. Data were analyzed using STATA v18, applying random-effects models to all analyses. Twelve studies involving 8321 patients were included, with most published in 2021-2024 (9/12). The pooled AUC of ML models for predicting LNM in CRC patients was 0.87 (95% CI: 0.82-0.91, I2:86.17) with a sensitivity of 78% (95% CI: 69-87%) and a specificity of 77% (95% CI: 64%-90%). In addition, when assessing the AUC reported by radiologists, both junior and senior radiologists had similar performance, significantly lower than the ML models. (P < 0.001). Subgroup analysis revealed higher AUCs in prospective studies (0.95, 95% CI: 0.87-1) compared to retrospective studies (0.85, 95% CI: 0.81-0.89) (P = 0.03). Studies without external validation exhibited significantly higher AUCs than those with external validation (P < 0.01). While there was no significant difference in AUC and sensitivity between the T1-T2 and T2-T4 stages, specificity was significantly higher in the T2-T4 stages than the low stages of T1 and T2 (95%, 95% CI: 92-98% vs. 61%, 95% CI: 44-78%; P < 0.01). ML models demonstrate strong potential for preoperative LNM staging and treatment planning in CRC, potentially reducing the need for additional surgeries and related health and financial burdens. Further prospective multicenter studies, with standardized reporting of algorithms, modality parameters, and LNM staging, are needed to validate these findings.

Investigating the Performance of Data-Driven Ensemble Machine Learning Models in Preliminary Designing Multi-Stage Friction Pendulum Bearings

Over the last few decades, the field of base isolation systems has demonstrated enormous advancements through the innovation of novel systems, and there has been an enhancement in the performance of structures equipped with isolation under conditions of moderate and severe seismic activity. One of the most important systems is the multi-stage friction pendulum, which offers a spectrum of effective pendulums across different regimes, thereby achieving enhanced capability for energy dissipation. The difficulty in designing these bearings at the preliminary stage comes from the fairly long process of trials and errors in selecting the properties of each sliding surface so that the required effective period, effective damping, and displacement capacities are achieved. This study investigates the performance of various data-driven ensemble machine learning models in directly designing multi-stage friction pendulum bearings. Within the context of the analysis, the most recent generation of friction pendulum that has been introduced to the literature, the quintuple friction pendulum, is used as the investigation employs a case to assess the dependability of the design strategy and the efficacy of a multi-output machine learning techniques. In general, the findings of this study have shown that machine learning models provide an accurate way to directly design quintuple friction pendulum isolators and attain the necessary properties of the isolators. The practicality of this research lies in highlighting how machine learning can be applied to considerably expedite the design and optimization of multi-stage friction pendulum bearings and in defining the most suitable machine learning technique to adopt in such tasks.

Pneumonia detection on chest X-rays from Xception-based transfer learning and logistic regression.

Pneumonia is a dangerous disease that kills millions of children and elderly patients worldwide every year. The detection of pneumonia from a chest x-ray is perpetrated by expert radiologists. The chest x-ray is cheaper and is most often used to diagnose pneumonia. However, chest x-ray-based diagnosis requires expert radiologists which is time-consuming and laborious. Moreover, COVID-19 and pneumonia have similar symptoms which leads to false positives. Machine learning-based solutions have been proposed for the automatic prediction of pneumonia from chest X-rays, however, such approaches lack robustness and high accuracy due to data imbalance and generalization errors. This study focuses on elevating the performance of machine learning models by dealing with data imbalanced problems using data augmentation. Contrary to traditional machine learning models that required hand-crafted features, this study uses transfer learning for automatic feature extraction using Xception and VGG-16 to train classifiers like support vector machine, logistic regression, K nearest neighbor, stochastic gradient descent, extra tree classifier, and gradient boosting machine. Experiments involve the use of hand-crafted features, as well as, transfer learning-based feature extraction for pneumonia detection. Performance comparison using Xception and VGG-16 features suggest that transfer learning-based features tend to show better performance than hand-crafted features and an accuracy of 99.23% can be obtained for pneumonia using chest X-rays.

Hybrid physically based and machine learning model to enhance high streamflow prediction

ABSTRACT Despite the significant performance of machine learning models for streamflow prediction, their precision for poorly represented data is reduced. This is a concern for flood mitigation purposes where high streamflow values are the most relevant but scarce. Consequently, this study proposes a methodology to create a hybrid model to mitigate the accuracy reduction of a standalone machine learning model in high streamflow values. The hybrid model combines a surrogate model that reproduces a physically based model with a model to estimate its residuals employing the Random Forest algorithm. The hybrid model reaches a root mean squared error reduction of 23% and 33% in the study catchments for values over a three-year return period compared to a standalone machine learning model. The percentage bias decreases by more than 70% from values over a 1.5-year return period. Moreover, the hybrid model has shown close predictions of values higher than the training set.

A novel importance-guided particle swarm optimization based on MLP for solving large-scale feature selection problems

Feature selection is a crucial data preprocessing technique that effectively reduces the dataset size and enhances the performance of machine learning models. Evolutionary computation (EC) based feature selection has become one of the most important parts of feature selection methods. However, the performance of existing EC methods significantly decrease when dealing with datasets with thousands of dimensions. To address this issue, this paper proposes a novel method called importance-guided particle swarm optimization based on MLP (IGPSO) for feature selection. IGPSO utilizes a two stage trained neural network to learn a feature importance vector, which is then used as a guiding factor for population initialization and evolution. In the two stage of learning, the positive samples are used to learn the importance of useful features while the negative samples are used to identify the invalid features. Then the importance vector is generated combining the two category information. Finally, it is used to replace the acceleration factors and inertia weight in original binary PSO, which makes the individual acceleration factor and social acceleration factor are positively correlated with the importance values, while the inertia weight is negatively correlated with the importance value. Further more, IGPSO uses the flip probability to update the individuals. Experimental results on 24 datasets demonstrate that compared to other state-of-the-art algorithms, IGPSO can significantly reduce the number of features while maintaining satisfactory classification accuracy, thus achieving high-quality feature selection effects. In particular, compared with other state-of-the-art algorithms, there is an average reduction of 0.1 in the fitness value and an average increase of 6.7% in classification accuracy on large-scale datasets.

Comparison and analysis of denoising method in TBM key tunnelling data

ABSTRACT Environmental vibrations and other factors cause significant random errors in raw data collected during TBM tunnelling. Such errors can reach up to 25% and will severely reduce the performance of data mining and machine learning model. To improve the data quality and find a suitable denoising method for TBM data, this study first proposed two assurance indicators to evaluate the effectiveness of data denoising, such as distortion degree (θ) and denoising magnitude (λ). Then, the effects of the data denoising schemes were compared. The results demonstrate that: (1) The Convlove19 scheme has better performance and is recommended for the denoising processing for TBM data. It can preserve the statistical characteristics of the raw data to the maximum extent possible and can improve the performance of the machine learning model by 15∼25%. (2) “Boring cycle segmentation after denoising strategy” (SAD) has better performance than “Boring cycle segmentation before denoising strategy” (SBD) and is recommended. The findings in this study can help with TBM data cleaning and quality improvement, as well as provide a reference for noise reduction of other data with similar features.

Data-driven prediction of flow fields in a needle-ring-net electrohydrodynamic pump system

In this paper, a data-mechanism hybrid modeling method for efficiently obtaining an electrohydrodynamic flow field is proposed. First, a backpropagation (BP) model with high accuracy is trained to get the value of essential parameter q0 for the mechanism simulation of flow fields. Subsequently, the mechanism model is used to generate a database for flow field reconstruction. Three machine learning algorithms, namely, BP neural network, random forest regression (RFR), and convolutional neural network (CNN), are employed to predict and reconstruct the flow behaviors of a needle-ring-net electrohydrodynamic pump. The RFR model demonstrates higher accuracy and precision in predicting velocity and pressure in the flow field compared to the BP and CNN models. The use of machine learning models for flow field prediction can significantly reduce the computational time while maintaining the computational accuracy. Additionally, an analysis assessing the impact of varying dataset sizes on the prediction accuracy of the model is conducted. The results indicate that the size of the dataset significantly influences the model predictive performance. Specifically, larger datasets are suggested to enhance both the accuracy and the generalization capabilities of the model. This observation highlights the critical role of dataset size in optimizing the performance of machine learning models for predictive tasks in engineering applications. These results offer important references for improving the design and optimization of electrohydrodynamic pumps.

Blending is all you need: Data-centric ensemble synthetic data

Deep generative models have gained increasing popularity, particularly in fields such as natural language processing and computer vision. Recently, efforts have been made to extend these advanced algorithms to tabular data. While generative models have shown promising results in creating synthetic data, their high computational demands and the need for careful parameter tuning present significant challenges. This study investigates whether a collective integration of refined synthetic datasets from multiple models can achieve comparable or superior performance to that of a single, large generative model. To this end, we developed a Data-Centric Ensemble Synthetic Data model, leveraging principles of ensemble learning. Our approach involved a data refinement process applied to various synthetic datasets, systematically eliminating noise and ranking, selecting, and combining them to create an augmented, high-quality synthetic dataset. This approach improved both the quantity and quality of the data. Central to this process, we introduced the Ensemble k-Nearest Neighbors with Centroid Displacement (EKCD) algorithm for noise filtering, alongside a density score for ranking and selecting data. Our experiments confirmed the effectiveness of EKCD in removing noisy synthetic samples. Additionally, the ensemble model based on the refined synthetic data substantially enhanced the performance of machine learning models, sometimes even outperforming that of real data.

Electrocardiogram sampling frequency for optimal performance of complexity analysis and machine learning models: Discrimination between subjects with and without paroxysmal atrial fibrillation using sinus rhythm electrocardiograms

BackgroundCurrent clinical practice to diagnose atrial fibrillation (AF) requires repeated, episodic monitoring and significantly underperform in their ability to detect AF episodes. ObjectiveThere is therefore potential for artificial intelligence based methods to assist in the detection of AF. Better understanding the optimal parameters for this detection can potentially improve the sensitivity in detecting AF. MethodTen second, 12-lead electrocardiogram signals were analysed using complexity algorithms combined with machine learning techniques to predict patients who had a previously detected AF episode but had since returned to normal sinus rhythm. An investigation was performed into the impact of sampling frequency of the electrocardiogram signal on the accuracy of the machine learning models used. ResultsUsing a single complexity algorithm showed a peak accuracy of 0.69 when using signals sampled at 125Hz. In particular it was noted that improved accuracy occurred when using lead V6 compared to other available leads. ConclusionBased on these results there is potential for 12-lead electrocardiogram signals to be recorded at 125Hz as standard and used in conjunction with complexity analysis to aid in the detection of subjects with AF.

Reconstructing MODIS normalized difference snow index product on Greenland ice sheet using spatiotemporal extreme gradient boosting model

The spatiotemporally continuous data of normalized difference snow index (NDSI) are key to understanding the mechanisms of snow occurrence and development as well as the patterns of snow distribution changes. However, the presence of clouds, particularly prevalent in polar regions such as the Greenland Ice Sheet (GrIS), introduces a significant number of missing pixels in the MODIS NDSI daily data. To address this issue, this study proposes the utilization of a spatiotemporal extreme gradient boosting (STXGBoost) model generate a comprehensive NDSI dataset. In the proposed model, various input variables are carefully selected, encompassing terrain features, geometry-related parameters, and surface property variables. Moreover, the model incorporates spatiotemporal variation information, enhancing its capacity for reconstructing the NDSI dataset. Verification results demonstrate the efficacy of the STXGBoost model, with a coefficient of determination of 0.962, root mean square error of 0.030, mean absolute error of 0.011, and negligible bias (0.0001). Furthermore, simulation comparisons involving missing data and cross-validation with Landsat NDSI data illustrate the model’s capability to accurately reconstruct the spatial distribution of NDSI data. Notably, the proposed model surpasses the performance of traditional machine learning models, showcasing superior NDSI predictive capabilities. This study highlights the potential of leveraging auxiliary data to reconstruct NDSI in GrIS, with implications for broader applications in other regions. The findings offer valuable insights for the reconstruction of NDSI remote sensing data, contributing to the further understanding of spatiotemporal dynamics in snow-covered regions.

A three-phase framework for mapping barriers to blockchain adoption in sustainable supply chain

PurposeBlockchain technology is one of the major contributors to supply chain sustainability because of its inherent features. However, its adoption rate is relatively low due to reasons such as the diverse barriers impeding blockchain adoption. The purpose of this study is to identify blockchain adoption barriers in sustainable supply chain and uncovers their interrelationships.Design/methodology/approachA three-phase framework that combines machine learning (ML) classifiers, BORUTA feature selection algorithm, and Grey-DEMATEL method. From the literature review, 26 potential barriers were identified and evaluated through the performance of ML models with accuracy and f-score.FindingsThe findings reveal that feature selection algorithm detected 15 prominent barriers, and random forest (RF) classifier performed with the highest accuracy and f-score. Moreover, the performance of the RF increased by 2.38% accuracy and 2.19% f-score after removing irrelevant barriers, confirming the validity of feature selection algorithm. An RF classifier ranked the prominent barriers and according to ranking, financial constraints, immaturity, security, knowledge and expertise, and cultural differences resided at the top of the list. Furthermore, a Grey-DEMATEL method is employed to expose interrelationships between prominent barriers and to provide an overview of the cause-and-effect group.Practical implicationsThe outcome of this study can help industry practitioners develop new strategies and plans for blockchain adoption in sustainable supply chains.Originality/valueThe research on the adoption of blockchain technology in sustainable supply chains is still evolving. This study contributes to the ongoing debate by exploring how practitioners and decision-makers adopt blockchain technology, developing strategies and plans in the process.

Predictive performance of machine learning models for kidney complications following coronary interventions: a systematic review and meta-analysis.

Acute kidney injury (AKI) and contrast-induced nephropathy (CIN) are common complications following percutaneous coronary intervention (PCI) or coronary angiography (CAG), presenting significant clinical challenges. Machine learning (ML) models offer promise for improving patient outcomes through early detection and intervention strategies. A comprehensive literature search following PRISMA guidelines was conducted in PubMed, Scopus, and Embase from inception to June 11, 2024. Study characteristics, ML models, performance metrics (AUC, accuracy, sensitivity, specificity, precision), and risk-of-bias assessment using the PROBAST tool were extracted. Statistical analysis used a random-effects model to pool AUC values, with heterogeneity assessed via the I2 statistic. From 431 initial studies, 14 met the inclusion criteria. Gradient Boosting Machine (GBM) and Support Vector Machine (SVM) models showed the highest pooled AUCs of 0.87 (95% CI: 0.82-0.92) and 0.85 (95% CI: 0.80-0.90), respectively, with low heterogeneity (I2 < 30%). Random Forest (RF) had a similar AUC of 0.85 (95% CI: 0.78-0.92) but significant heterogeneity (I2 > 90%). Multilayer perceptron (MLP) and XGBoost models had moderate pooled AUCs of 0.79 (95% CI: 0.74-0.84) with high heterogeneity. RF showed strong accuracy (0.83, 95% CI: 0.70-0.96), while SVM had balanced sensitivity (0.69, 95% CI: 0.63-0.75) and specificity (0.73, 95% CI: 0.60-0.86). Age, serum creatinine, left ventricular ejection fraction, and hemoglobin consistently influenced model efficacy. GBM and SVM models, with robust AUCs and low heterogeneity, are effective in predicting AKI and CIN post-PCI/CAG. RF, MLP, and XGBoost, despite competitive AUCs, showed considerable heterogeneity, emphasizing the need for further validation.

Evaluating The Performance of Machine Learning Models in Audit Opinion Prediction – A Study in Vietnam

This study investigates the effectiveness of machine learning models in predicting audit opinions using a dataset from the FiinPro-X platform, comprising 9,783 audited consolidated financial statements from public companies listed on Vietnamese stock exchanges from 2016 to 2023. The dataset spans various industries, excluding banks and financial institutions, and focuses on identifying key financial, non-financial, and qualitative variables that influence audit opinions. Six supervised learning algorithms were applied—Logistic Regression, K-Nearest Neighbors (KNN), Decision Trees, Random Forests, Support Vector Machines (SVM), and Naive Bayes—evaluated based on their ability to predict both fully acceptable (unqualified) and non-fully acceptable audit opinions. All data processing and model training were implemented in a Python environment. The Random Forest model demonstrated the best overall performance, achieving an accuracy of 0.868 and an AUC-ROC of 0.87, though its F1 score for predicting non-fully acceptable audit opinions was lower (0.585). This suggests that while machine learning models can improve prediction accuracy, challenges remain in handling imbalanced data and non-linear relationships among input variables. The study also reduced the number of features by 30%, improving the models’ performance. Future research should further refine data and feature construction processes to ensure comparability and practical applicability.

Data-driven AI for the automated classification of the isothermal heat-treated thermal barrier coatings using pulsed infrared thermography

Abstract Thermal barrier coating (TBC) is a multilayer coating applied to metallic structures exposed to high temperatures, such as aero-engine parts and gas turbine blades. These thermally insulating materials are used to extend the component life by minimizing the high-temperature exposure of structural components. As a result of the high-temperature oxidation tests, a thermally grown oxide (TGO) layer formed at the interface between the ceramic topcoat layer and the bond coat layer. The primary cause of TBC performance decline and structure failure is the growth of the oxide layer. Hence, it is crucial to regularly monitor the formation of the oxide layer and the degradation of coatings. This can be achieved by developing coating-life prediction models, which are vital for quality control, health monitoring of TBCs, maintenance decision-making, and the prevention of catastrophic failures. Existing methods for life estimation, which are dependent on empirical methods and microstructural analysis, often fail due to various material compositions and service conditions. This study proposes an automated classification of iso-thermal heat-treated TBC samples using temperature data, which helps in predicting the TBC life and monitoring the TBC degradation. TBC-coated samples are isothermal heat-treated at 1000 °C, and the initial growth of TGO is monitored using a non-destructive thermal imaging technique. Identifying distinctive features in each class corresponding to layer thickness measurement and service hours is challenging and time-consuming. So, we integrate data-driven AI models and feature extraction techniques to interpret complex thermal patterns. The performance of deep learning models and classical machine learning models demonstrates that our method can achieve maximum classification accuracy and F1-score of 89.2% and 89.0%, respectively. This automated classification framework promises a powerful tool for predictive maintenance and remaining useful lifetime estimation of hot section components reliant on TBCs.

Enhancing Heart Disease Prediction through Machine Learning: A Comparative Analysis of Algorithm Generalization across Datasets with Various Distributions

Heart disease, due to its high prevalence and mortality, remains a key area of global research. Although traditional diagnostic methods are effective, they are often invasive and time-consuming, highlighting the need for non-invasive, AI-based approaches. A significant challenge in real-world applications is ensuring model generalization across different datasets, particularly when the datasets are small. In this study, the performance of machine learning models, including Decision Tree, Random Forest, Support Vector Machine (SVM), and Multi-Layer Perceptron (MLP), was evaluated on two distinct heart disease datasets with different distributions and relatively small sizes. Two datasets with varying distributions were used for training and testing, with the primary focus on assessing model generalization in crossdataset applications. It is shown in the results that, while the Decision Tree model performed best after hyperparameter tuning, the improvements in Random Forest and MLP were limited, and SVM exhibited a decline in performance after tuning in the cross-dataset task. It was found that grid search tuning has limitations in cross-dataset scenarios, especially with small datasets, where complex models are prone to overfitting. The study demonstrates that, with smaller datasets, simpler models like Decision Trees often adapt better to different datasets. Furthermore, transfer learning and domain adaptation techniques are suggested as crucial for improving model generalization. Future research should focus on employing these techniques to enhance the robustness and accuracy of heart disease prediction models across diverse datasets.

A novel multi-model feature generation technique for suicide detection

Automated expert systems (AES) analyzing depression-related content on social media have piqued the interest of researchers. Depression, often linked to suicide, requires early prediction for potential life-saving interventions. In the conventional approach, psychologists conduct patient interviews or administer questionnaires to assess depression levels. However, this traditional method is plagued by limitations. Patients might not feel comfortable disclosing their true feelings to psychologists, and counselors may struggle to accurately predict situations due to limited data. In this context, social media emerges as a potentially valuable resource. Given the widespread use of social media in daily life, individuals often express their nature and mental state through their online posts. AES can efficiently analyze vast amounts of social media content to predict depression levels in individuals at an early stage. This study contributes to this endeavor by proposing an innovative approach for predicting suicide risks using social media content and machine learning techniques. A novel multi-model feature generation technique is employed to enhance the performance of machine learning models. This technique involves the use of a feature extraction method known as term frequency-inverse document frequency (TF-IDF), combined with two machine learning models: logistic regression (LR) and support vector machine (SVM). The proposed technique calculates probabilities for each sample in the dataset, resulting in a new feature set referred to as the probability-based feature set (ProBFS). This ProBFS is compact yet highly correlated with the target classes in the dataset. The utilization of concise and correlated features yields significant outcomes. The SVM model achieves an impressive accuracy score of 0.96 using ProBFS while maintaining a low computational time of 5.63 seconds even when dealing with extensive datasets. Furthermore, a comparison with state-of-the-art approaches is conducted to demonstrate the significance of the proposed method.

Generative adversarial networks (GANs) to produce synthetic 12-lead electrocardiogram signals for specific and rare diseases: a novel, powerful tool towards clinical advancements

Abstract Background Remarkable technological advancements in electrocardiogram (ECG) analysis face challenges due to insufficiently diverse and extensive real-world datasets. This limitation hampers both the development of sophisticated deep learning (DL) models for ECG analysis and the exposure of new physicians to substantial training examples. Purpose To leverage Generative Adversarial Networks (GANs), a novel AI algorithm type, to generate synthetic yet realistic ECG signals resembling specific diseases. Methods We initially employed a general-purpose GAN on more than 21,000 ECG samples from the PTB-XL database, encompassing normal and abnormal ECGs across various diseases. The GAN outputs served as inputs for training "DeceptionECG", a novel GAN model inspired by the InceptionTime architecture (code available online). DeceptionECG focused on generating disease-specific ECG signals, trained on annotated ECGs validated by two electrophysiologists. Post-training, we generated synthetic ECGs for common diseases like atrial fibrillation (AF) and anterior ST-segment elevation myocardial infarction (STEMI), as well as rarer conditions like Wolf-Parkinson White (WPW) syndrome. Authentication involved an independent discriminator DL model and two cardiologists who were asked to distinguish real from synthetic ECGs. Additionally, a DL model, a variant of the InceptionTime architecture, discerned specific diseases in the ECG signals to assess their realism. Finally, we trained a DL model for detecting abnormalities in ECGs with the augmented dataset to evaluate potential performance improvement. Results We generated sets of 5,000 synthetic ECGs for various abnormalities, at a rate of 2,500 12-lead ECG signals per minute. The generated ECGs were indistinguishable from real ones, evidenced by low accuracy (Accuracy, ACC: 48.9%) of the discriminator model and two clinicians (ACC: 56% and 51%), on test sets with anterior STEMI. A DL model trained on real ECGs accurately differentiated between synthetic normal ECGs and those hosting the target disease in nearly 9 out of 10 cases (ACC: 88.7%). Furthermore, the employed DL model detector trained on the augmented dataset, encompassing both real (n=1,589) and synthetic (n=5,000) ECGs, outperformed its counterpart trained solely on real ECGs, when tested on a 1,000-sample set of real signals (ACC: 94.6% vs. 89.3%). Conclusion Our GAN-based approach showcases the potential of generating high-quality synthetic ECGs tailored for specific diseases, overcoming data scarcity challenges, and enhancing the training and performance of machine learning models. Additionally, it holds promise in advancing medical professional training by providing diverse signals with distinct "disease stamps" at massive amounts, addressing privacy concerns associated with extensive data exposure. Further testing is warranted for the applicability of the approach, especially in exceedingly rare conditions with limited available training data.Project workflow and outcomesGAN-generated ECGs: Which is fake?