PureFlow: An Unsupervised Autoencoder-Based Dataset Purification Framework for Malicious Traffic Detection

Cardiovascular diseases (CVDs) remain a leading cause of global mortality, underscoring the urgent need for early and accurate diagnosis. Electrocardiograms (ECGs) are a widely used, noninvasive tool for detecting cardiac abnormalities, but manual interpretation can be time- consuming and prone to error. This study explores advanced machine learning (ML) and deep learning (DL) methods to improve the detection of CVDs from ECG images, leveraging their ability to analyze complex patterns with high precision. Models such as Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), and ensemble frameworks are utilized, alongside traditional ML models like XGBoost, CatBoost, and LightGBM. The performance of these models is evaluated based on precision, recall, F1-score, and accuracy. Results show that DL models, particularly an ensemble of CNN and ViT, achieve superior accuracy , outperforming standalone ML methods. Furthermore, an ensemble of XGBoost, CatBoost, and LightGBM achieves competitive results, while Optuna-tuned XGBoost further enhances ML performance. The findings highlight the potential of combining ML and DL techniques for robust and reliable CVD detection in ECG images. This research demonstrates the promise of AI-driven solutions in advancing medical diagnostics and supports the integration of automated systems for early detection of cardiovascular diseases.

Deep learning (DL) and machine learning (ML) methods have recently contributed to the advancement of models in the various aspects of prediction, planning, and uncertainty analysis of smart cities and urban development. This paper presents the state of the art of DL and ML methods used in this realm. Through a novel taxonomy, the advances in model development and new application domains in urban sustainability and smart cities are presented. Findings reveal that five DL and ML methods have been most applied to address the different aspects of smart cities. These are artificial neural networks; support vector machines; decision trees; ensembles, Bayesians, hybrids, and neuro-fuzzy; and deep learning. It is also disclosed that energy, health, and urban transport are the main domains of smart cities that DL and ML methods contributed in to address their problems.

Feature mining for encrypted malicious traffic detection with deep learning and other machine learning algorithms

Mental disorders are a significant medical and social issue globally. Currently, approximately 970 million individuals suffer from mental disorders, with over 300 million diagnosed with depression or bipolar disorder. Recently, there has been significant advancement in digital technologies, particularly in artificial intelligence, encompassing machine learning and deep learning. Given the growing interest in their use in psychiatry and the need to develop new approaches to psychiatric care. This review explores the current and promising directions for the application of artificial intelligence technologies in clinical practice, focusing on patients with depression and bipolar disorder. A literature search was conducted from January to February 2024 in the databases PubMed, Google Scholar, and eLibrary using the following keywords: «психиатрия» ("psychiatry"), «психическое здоровье» ("mental health"), «психическое расстройство» ("psychiatric disorder"), «депрессия» ("depression"), «депрессивный эпизод» ("depressive episode"), «рекуррентное депрессивное расстройство» ("recurrent brief depression"), «биполярное расстройство» ("bipolar disorder"), «машинное обучение» ("machine learning"), «глубокое обучение» ("deep learning"), «искусственный интеллект» ("artificial intelligence"); "psychiatry", "mental health", "psychiatric disorder", "depression", "depressive episode", "major depressive disorder", "bipolar disorder", "machine learning", "deep learning", "artificial intelligence". Studies on the use of artificial intelligence technologies in patients with depression and bipolar disorders and review articles discussing the difficulties of their application in psychiatry were excluded. Publications in Russian and English in the past 10 years were selected. The most commonly used machine learning models for diagnosing patients with affective disorders utilize neuroimaging data (primarily magnetic resonance imaging and electroencephalography), text, audio, and video data and data from electronic devices, molecular-genetic markers, and clinical indicators. The models were trained using mono- or multimodal datasets. Notably, many of the reviewed studies have significant limitations, making the implementation of artificial intelligence technologies in clinical practice challenging. These include small sample sizes, low representativeness and standardization, inclusion of “noise” and correlated variables, and absence of validation using independent datasets. Studies on machine learning methods have demonstrated promising results in the early diagnosis of affective episodes and in predicting treatment responses. However, their clinical application is limited, owing to insufficient validation. Well-designed prospective cohort studies and the creation of extensive, high-quality datasets and models capable of uncovering new relationships between variables are required to address this limitation.

Photometric redshifts (photo- z ’s) are crucial for the cosmology, galaxy evolution, and transient science drivers of next-generation imaging facilities like the Euclid Mission, the Vera C. Rubin Observatory, and the Nancy Grace Roman Space Telescope. Previous work has shown that image-based deep learning photo- z methods produce smaller scatter than photometry-based classical machine learning (ML) methods on the Sloan Digital Sky Survey (SDSS. Main Galaxy Sample, a test bed photo- z dataset. However, global assessments can obscure local trends. To explore this possibility, we used a self-organizing map (SOM) to cluster SDSS galaxies based on their ugriz colors. Deep learning methods achieve lower photo- z scatter than classical ML methods for all SOM cells. The fractional reduction in scatter is roughly constant across most of color space with the exception of the most bulge-dominated and reddest cells where it is smaller in magnitude. Interestingly, classical ML photo- z ’s suffer from a significant color-dependent attenuation bias, where photo- z ’s for galaxies within an SOM cell are systematically biased towards the cell’s mean spectroscopic redshift and away from extreme values, which is not readily apparent when all objects are considered. In contrast, deep learning photo- z ’s suffer from very little color-dependent attenuation bias. The increased attenuation bias for classical ML photo- z methods is the primary reason why they exhibit larger scatter than deep learning methods. This difference can be explained by the deep learning methods weighting redshift information from the individual pixels of a galaxy image more optimally than integrated photometry.

Machine learning methods, including Deep learnings are popular for data processing. Deep learning methods have shown great promise in applied in natural language processing (NLP) tasks. Text classification is a supervised machine learning task that involves labelled documents to train classifier. Previous works involving machine learning and deep learning methods for text classification have been tested with relatively small- sized data instances. In this paper, we compared the performance of the machine leaning and deep learning algorithms in text classification task. This paper also studied, explored and compared the scalability of these methods with respect to bigger data instances. We used support vector machines (SVM), Logistic regression, Random forest and Naive Bayes Machine leaning algorithms and convolutional neural network (CNN) deep learning method. The task involved a multi-class classification problem involving six (6) classes consisting of six thousand (6,000) data instances with an average of 20 sentences in each data instance. The CNN deep leaning algorithm outperformed all the machine learning algorithms, achieving an accuracy of over 85%. This is because the filter weights are leaning are updated in backward propagation in each epoch, hence, this result in better result compared to the traditional methods.

<p class="0abstract">The research aimed to conduct an extensive study of machine learning and deep learning methods in cybersecurity. To accomplish the objectives, the research carried out a qualitative study based on secondary data collection to review the available studies and literature. The research has examined three machine learning methods and three deep learning methods to study the most popular techniques used in cybersecurity. During the research, the working mechanism of each method was studied along with their strengths and weaknesses. Also, a comparative discussion has been made to examine the most effective method for cybersecurity. Limitations in the current literature were also identified, and future direction is also given to target and develop the weak areas of machine learning and deep learning methods.<em> </em><strong></strong></p>

Domain name generation algorithm (DGA) classification is an essential but challenging problem. Both feature-extracting machine learning (ML) methods and deep learning (DL) models such as convolutional neural networks and long short-term memory have been developed. However, the performance of these approaches varies with different types of DGAs. Most features in the ML methods can characterize random-looking DGAs better than word-looking DGAs. To improve the classification performance on word-looking DGAs, subword tokenization is employed for the DL models. Our experimental results proved that the subword tokenization can provide excellent classification performance on the word-looking DGAs. We then propose an integrated scheme that chooses an appropriate method for DGA classification depending on the nature of the DGAs. Results show that the integrated scheme outperformed existing ML and DL methods, and also the subword DL methods.

Background: Customer churn significantly impacts business revenues. Machine Learning (ML) and Deep Learning (DL) methods are increasingly adopted to predict churn, yet a systematic synthesis of recent advancements is lacking. Objectives: This systematic review evaluates ML and DL approaches for churn prediction, identifying trends, challenges, and research gaps from 2020 to 2024. Data Sources: Six databases (Springer, IEEE, Elsevier, MDPI, ACM, Wiley) were searched via Lens.org for studies published between January 2020 and December 2024. Study Eligibility Criteria: Peer-reviewed original studies applying ML/DL techniques for churn prediction were included. Reviews, preprints, and non-peer-reviewed works were excluded. Methods: Screening followed PRISMA 2020 guidelines. A two-phase strategy identified 240 studies for bibliometric analysis and 61 for detailed qualitative synthesis. Results: Ensemble methods (e.g., XGBoost, LightGBM) remain dominant in ML, while DL approaches (e.g., LSTM, CNN) are increasingly applied to complex data. Challenges include class imbalance, interpretability, concept drift, and limited use of profit-oriented metrics. Explainable AI and adaptive learning show potential but limited real-world adoption. Limitations: No formal risk of bias or certainty assessments were conducted. Study heterogeneity prevented meta-analysis. Conclusions: ML and DL methods have matured as key tools for churn prediction, yet gaps remain in interpretability, real-world deployment, and business-aligned evaluation. Systematic Review Registration: Registered retrospectively in OSF.

Accurate and effective classification of plant diseases is critical for increasing yield and quality in agricultural production, minimizing economic losses through early detection of diseases, and implementing sustainable agriculture approaches. In this study, an approach for the detection and classification of plant leaf diseases is presented. We compare the performance of machine learning and deep learning-based models, and we use GAN-based data synthesis methods on a dataset we created to improve the model performance. For machine learning methods, ResNet-based feature extraction is performed and XGBoost, Random Forest, SVM, and InceptionV3 models are evaluated. In contrast, AlexNet, VGG16, VGG19, DenseNet, and ResNet models are examined within the scope of deep learning. The study was analyzed in three classes Phytophthora Infestans, Potassium Deficiency, and Healthy, and tested on data obtained from 21 different plant species. According to the model performances obtained, the deep learning-based ResNet model showed the highest success in all performance metrics and achieved 99% accuracy, showing superior performance compared to other methods. In the study, a comprehensive evaluation of multiple classification, GAN-based data synthesis, machine learning, and deep learning models was carried out. A valuable contribution was made to the existing studies in the literature.

Emotion classification is a research field that aims to detect the emotions in a text using machine learning methods. In traditional machine learning (TML) methods, feature engineering processes cause the loss of some meaningful information, and classification performance is negatively affected. In addition, the success of modelling using deep learning (DL) approaches depends on the sample size. More samples are needed for Turkish due to the unique characteristics of the language. However, emotion classification data sets in Turkish are quite limited. In this study, the pretrained language model approach was used to create a stronger emotion classification model for Turkish. Well-known pretrained language models were fine-tuned for this purpose. The performances of these fine-tuned models for Turkish emotion classification were comprehensively compared with the performances of TML and DL methods in experimental studies. The proposed approach provides state-of-the-art performance for Turkish emotion classification.

SummaryAccurate prediction of the solubility of chemical substances in solvents remains a challenge. The sparsity of high-quality solubility data is recognized as the biggest hurdle in the development of robust data-driven methods for practical use. Nonetheless, the effects of the quality and quantity of data on aqueous solubility predictions have not yet been scrutinized. In this study, the roles of the size and the quality of data sets on the performances of the solubility prediction models are unraveled, and the concepts of actual and observed performances are introduced. In an effort to curtail the gap between actual and observed performances, a quality-oriented data selection method, which evaluates the quality of data and extracts the most accurate part of it through statistical validation, is designed. Applying this method on the largest publicly available solubility database and using a consensus machine learning approach, a top-performing solubility prediction model is achieved.

Accurate personal exposure assessment to air pollution is crucial in environmental epidemiological studies, but it remains challenging due to substantial differences in pollutant concentrations between indoor and outdoor environments. Indirect assessment methods estimate personal exposure by incorporating microenvironment-specific pollutant concentrations and personal time-activity patterns, while requiring participants to record detailed time-location diaries. This study aims to develop microenvironment classification models using global positioning system (GPS) tracking data. We used data from the Korean Air pollutant EXposure (KAPEX) model project, which reflects seasonal and daily activity patterns of the Korean urban population. Three microenvironment categorizations were defined: Two-level (indoor and outdoor), Three-level (indoor, walk, and transit), and Four-level (home and workplace, other indoor, walk, and transit), where the outdoor category is subdivided into walk and transit in Three- and Four-level. Classification models were developed for each categorization using various machine learning (ML) and deep learning (DL) methods, incorporating both individual mobility patterns and GPS signal quality information. Important variables contributing to classification accuracy were identified using interpretable ML or DL methods. Random forest achieved the highest test AUROC of 0.963 and 0.958 for the Two-level and Three-level categorization, respectively, while boosting achieved the highest test AUROC of 0.918 for the Four-level categorization. Variables related to both mobility patterns and GPS signal quality were found to enhance classification accuracy. These findings highlight that incorporating both mobility patterns and signal quality into ML models significantly improves classification accuracy. Microenvironment classification models were developed using GPS tracking data to improve personal exposure assessment. Both mobility patterns and GPS signal quality were used to enhance model accuracy. Random forest and boosting outperformed the deep neural network across various outcome categorizations. SHAP values were used to identify important variables contributing to model performance.

The global burden of diabetes mellitus (DM) continues to escalate, posing significant challenges to healthcare systems worldwide. This study compares machine learning (ML) and deep learning (DL) methods, their hybrids, and ensemble strategies for predicting the health outcomes of diabetic patients. This work aims to find the best solutions that strike a balance between computational efficiency and accurate prediction. The study systematically assessed a range of predictive models, including sophisticated DL techniques and conventional ML algorithms, based on computational efficiency and performance indicators. The study assessed prediction accuracy, processing speed, scalability, resource consumption, and interpretability using publicly accessible diabetes datasets. It methodically evaluates the selected models using key performance indicators (KPIs), training times, and memory usage. AdaBoost had the highest F1-score (0.74) on PIMA-768, while RF excelled on PIMA-2000 (~0.73). An RNN led the 3-class BRFSS survey (0.44), and a feed-forward DNN excelled on the binary BRFSS subset (0.45), while RF also achieved perfect accuracy on the EMR dataset (1.00) confirming that model performance is tightly coupled to each dataset’s scale, feature mix and label structure. The results highlight how lightweight, interpretable ML and DL models work in resource-constrained environments and for real-time health analytics. The study also compares its results with existing prediction models, confirming the benefits of selected ML approaches in enhancing diabetes-related medical outcomes that are substantial for practical implementation, providing a reliable and efficient framework for automated diabetes prediction to support initiative-taking disease management techniques and tailored treatment. The study concludes the essentiality of conducting a thorough assessment and validation of the model using current institutional datasets as this enhances accuracy, security, and confidence in AI-assisted healthcare decision-making.

Peptides have shown increasing advantages and significant clinical value in drug discovery and development. With the development of high-throughput technologies and artificial intelligence (AI), machine learning (ML) methods for discovering new lead peptides have been expanded and incorporated into rational drug design. Predictions of peptide-protein interactions (PepPIs) and protein-protein interactions (PPIs) are both opportunities and challenges in computational biology, which will help to better understand the mechanisms of disease and provide the impetus for the discovery of lead peptides. This paper comprehensively reviews computational models for PepPI and PPI predictions. It begins with an introduction of various databases of peptide ligands and target proteins. Then it discusses data formats and feature representations for proteins and peptides. Furthermore, classical ML methods and emerging deep learning (DL) methods that can be used to train prediction models of PepPI and PPI are classified into four categories, and their advantages and disadvantages are analyzed. To assess the relative performance of different models, different validation protocols and evaluation indexes are discussed. The goal of this review is to help researchers quickly get started to develop computational frameworks using these integrated resources and eventually promote the discovery of lead peptides.