Data Analysis Algorithms Research Articles

INFLUENCE OF ANOMALOUS VALUES ON THE ACCURACY OF FORECASTING LOSSES IN DISTRIBUTION NETWORKS

The relevance of the study is determined by modern trends in managing the operation modes of distribution electrical networks using Smart Grid technologies, as well as the need to reduce the costs incurred by distribution system operators for purchasing electricity. Accurate load forecasting results at network nodes for different forecasting horizons are crucial for this purpose. Sudden changes in network topology can increase the errors in loss forecasts as a single time series, negatively impacting network management efficiency and increasing the costs of electricity procurement to cover losses. The study proposes using forecasting methods based on artificial neural networks for the calculation and prediction of electricity losses, along with a comparison of these methods. The calculations were performed using data from one of Ukraine's distribution system operators, and the test electrical network was adapted based on the CIGRE scheme for modeling electricity losses. Since the nodal load data contained gaps and anomalies, a two-step data analysis algorithm was employed using the DBSCAN clustering method for detection and correction. As a result of loss calculations based on cleaned data, the error was reduced threefold compared to calculations based on load factors. Applying data analysis methods and forecasting methods based on artificial neural networks significantly improves the accuracy of loss calculations and minimizes errors.

Low-Cost Approaches in Neuroscience to Teach Machine Learning Using a Cockroach Model.

In an effort to increase access to neuroscience education in underserved communities, we created an educational program that utilizes a simple task to measure place preference of the cockroach (Gromphadorhina portentosa) and the open-source free software, SLEAP Estimates Animal Pose (SLEAP) to quantify behavior. Cockroaches (n = 18) were trained to explore a linear track for 2 min while exposed to either air, vapor, or vapor with nicotine from a port on one side of the linear track over 14 d. The time the animal took to reach the port was measured, along with distance traveled, time spent in each zone, and velocity. As characterizing behavior is challenging and inaccessible for nonexperts new to behavioral research, we created an educational program using the machine learning algorithm, SLEAP, and cloud-based (i.e., Google Colab) low-cost platforms for data analysis. We found that SLEAP was within a 0.5% margin of error when compared with manually scoring the data. Cockroaches were found to have an increased aversive response to vapor alone compared with those that only received air. Using SLEAP, we demonstrate that the x-y coordinate data can be further classified into behavior using dimensionality-reducing clustering methods. This suggests that the linear track can be used to examine nicotine preference for the cockroach, and SLEAP can provide a fast, efficient way to analyze animal behavior. Moreover, this educational program is available for free for students to learn a complex machine learning algorithm without expensive hardware to study animal behavior.Significance Statement This method demonstrates a novel utilization of machine learning using free cloud-based programming to analyze cockroach behavior in a linear track. This educational program can be implemented in the classroom as a low-cost tool to teach neuroscience and machine learning. Further, implementing these computational tools allows students to explore important questions in behavioral neuroscience, such as learning and memory, drug seeking, and exploratory locomotor behavior.

Smart Machine Learning Analysis and Blockchain Healthcare Platform

The Smart Machine Learning Medical Analysis and Blockchain Healthcare Platform (SMLMAP) represents a transformative approach to medical diagnostics and data management by integrating advanced machine learning algorithms with blockchain technology. This platform is designed to accurately diagnose seven critical diseases, achieving impressive diagnostic outcomes across various conditions, including diabetes, breast cancer, heart disease, kidney disease, liver disease, malaria, and pneumonia. SMLMAP employs Random Forest algorithms for text data analysis and Convolutional Neural Networks (CNNs) for image data, ensuring rapid and precise diagnostic results. In addition to enhancing diagnostic capabilities, SMLMAP addresses significant privacy and security concerns inherent in healthcare data management. The integration of blockchain technology facilitates secure, transparent, and tamper-proof storage of patient records, empowering individuals with greater control over their medical information. This dual approach not only improves the accuracy of diagnoses but also fosters trust between patients and healthcare providers by safeguarding sensitive data against unauthorized access. Key functionalities of SMLMAP include the ability to schedule doctor appointments conveniently through the platform, ensuring that patients can receive timely medical attention. Additionally, the application provides users with a weekly healthcare magazine, delivering valuable information on health tips, disease management, and wellness advice to promote informed health decisions. By leveraging the strengths of machine learning and blockchain, SMLMAP paves the way for a more efficient, accurate, and secure healthcare ecosystem. This innovative platform not only enhances the quality of medical diagnostics but also represents a significant step forward in addressing the critical challenges of data privacy and security in the healthcare sector. The findings underscore the potential of SMLMAP to revolutionize patient care and data management in the evolving landscape of healthcare technology.

Preliminary investigations into optimizing detector configuration for nondestructive assay of fresh nuclear fuel

Like other components of the nuclear fuel cycle, fresh nuclear fuel has potential safeguard and security concerns. Non-destructive assay (NDA) measurements need to be examined and made available to nuclear industry partners to decrease safeguard and oversight concerns for nuclear power plants and facilities. This study analyzed four proposed evaluation criteria: quantity of detectors required, speed and accuracy of detector verification, overall efficiency of neutron and gamma detections, and suitability of observables to advanced machine learning algorithms. Four different detectors have been arranged to be evaluated simultaneously: four Saint-Gobain 3He proportional counters, an Eljen EJ-309 liquid scintillator, a Saint-Gobain / Bicron NaI(Tl) scintillation detector, and a Cs2LiYCl6:Ce3+(CLYC) scintillation detector. Mock-up BWR and PWR fresh fuel configurations were developed by creating 4x4 and 3x3 natural uranium assemblies and aided in assessments for detector configurations. Additionally, 252Cf was included to increase the number of neutrons for evaluation. The results of this work will be used to develop a versatile, multi-detector system for fresh fuel verification. The exploration of detector types, data modalities, and their significance as input for advanced machine learning and data analysis algorithms is being explored to understand the obstacles and opportunities in the development of a consolidated low-cost, commercial-off-the-shelf system that nuclear power plants, facilities, and/or inspectors can use for non-destructive verification of fresh nuclear fuel and assembly templating for quick nuclear material inventory verification.

A big data analysis algorithm for massive sensor medical images.

Big data analytics for clinical decision-making has been proposed for various clinical sectors because clinical decisions are more evidence-based and promising. Healthcare data is so vast and readily available that big data analytics has completely transformed this sector and opened up many new prospects. The smart sensor-based big data analysis recommendation system has significant privacy and security concerns when using sensor medical images for suggestions and monitoring. The danger of security breaches and unauthorized access, which might lead to identity theft and privacy violations, increases when sending and storing sensitive medical data on the cloud. Our effort will improve patient care and well-being by creating an anomaly detection system based on machine learning specifically for medical images and providing timely treatments and notifications. Current anomaly detection methods in healthcare systems, such as artificial intelligence and big data analytics-intracerebral hemorrhage (AIBDA-ICH) and parallel conformer neural network (PCNN), face several challenges, including high resource consumption, inefficient feature selection, and an inability to handle temporal data effectively for real-time monitoring. Techniques like support vector machines (SVM) and the hidden Markov model (HMM) struggle with computational overhead and scalability in large datasets, limiting their performance in critical healthcare applications. Additionally, existing methods often fail to provide accurate anomaly detection with low latency, making them unsuitable for time-sensitive environments. We infer the extraction, feature selection, attack detection, and data collection and processing procedures to anticipate anomaly inpatient data. We transfer the data, take care of missing values, and sanitize it using the pre-processing mechanism. We employed the recursive feature elimination (RFE) and dynamic principal component analysis (DPCA) algorithms for feature selection and extraction. In addition, we applied the Auto-encoded genetic recurrent neural network (AGRNN) approach to identify abnormalities. Data arrival rate, resource consumption, propagation delay, transaction epoch, true positive rate, false alarm rate, and root mean square error (RMSE) are some metrics used to evaluate the proposed task.

Integrating artificial intelligence with smartphone-based imaging for cancer detection in vivo

Cancer is a major global health challenge, accounting for nearly one in six deaths worldwide. Early diagnosis significantly improves survival rates and patient outcomes, yet in resource-limited settings, the scarcity of medical resources often leads to late-stage diagnosis. Integrating artificial intelligence (AI) with smartphone-based imaging systems offers a promising solution by providing portable, cost-effective, and widely accessible tools for early cancer detection. This paper introduces advanced smartphone-based imaging systems that utilize various imaging modalities for in vivo detection of different cancer types and highlights the advancements of AI for in vivo cancer detection in smartphone-based imaging. However, these compact smartphone systems face challenges like low imaging quality and restricted computing power. The use of advanced AI algorithms to address the optical and computational limitations of smartphone-based imaging systems provides promising solutions. AI-based cancer detection also faces challenges. Transparency and reliability are critical factors in gaining the trust and acceptance of AI algorithms for clinical application, explainable and uncertainty-aware AI breaks the black box and will shape the future AI development in early cancer detection. The challenges and solutions for improving AI accuracy, transparency, and reliability are general issues in AI applications, the AI technologies, limitations, and potentials discussed in this paper are applicable to a wide range of biomedical imaging diagnostics beyond smartphones or cancer-specific applications. Smartphone-based multimodal imaging systems and deep learning algorithms for multimodal data analysis are also growing trends, as this approach can provide comprehensive information about the tissue being examined. Future opportunities and perspectives of AI-integrated smartphone imaging systems will be to make cutting-edge diagnostic tools more affordable and accessible, ultimately enabling early cancer detection for a broader population.

Dual assurance for healthcare and future education development: normalized assistance for low-income population in rural areas-evidence from the population identification.

This study aims to explore the relationship between healthcare and future education among the rural low-income population, using J City in Guangdong Province as the focal area. Addressing both healthcare and educational concerns, this research seeks to provide insights that can guide policy and support for this demographic. Utilizing big data analysis and deep learning algorithms, a targeted intelligent identification classification model was developed to accurately detect and classify rural low-income individuals. Additionally, a questionnaire survey methodology was employed to separately investigate healthcare and future education dimensions among the identified population. The proposed model achieved a population identification accuracy of 91.93%, surpassing other baseline neural network algorithms by at least 2.65%. Survey results indicated low satisfaction levels in healthcare areas, including medical resource distribution, medication costs, and access to basic medical facilities, with satisfaction rates below 50%. Regarding future education, issues such as tuition burdens, educational opportunity disparities, and accessibility challenges highlighted the concerns of rural low-income families. The high accuracy of the model demonstrates its potential for precise identification and classification of low-income populations. Insights derived from healthcare and education surveys reveal systemic issues affecting satisfaction and accessibility. This research thus provides a valuable foundation for future studies and policy development targeting rural low-income populations in healthcare and education.

Advances in Wearable Biosensors for Healthcare: Current Trends, Applications, and Future Perspectives.

Wearable biosensors are a fast-evolving topic at the intersection of healthcare, technology, and personalized medicine. These sensors, which are frequently integrated into clothes and accessories or directly applied to the skin, provide continuous, real-time monitoring of physiological and biochemical parameters such as heart rate, glucose levels, and hydration status. Recent breakthroughs in downsizing, materials science, and wireless communication have greatly improved the functionality, comfort, and accessibility of wearable biosensors. This review examines the present status of wearable biosensor technology, with an emphasis on advances in sensor design, fabrication techniques, and data analysis algorithms. We analyze diverse applications in clinical diagnostics, chronic illness management, and fitness tracking, emphasizing their capacity to transform health monitoring and facilitate early disease diagnosis. Additionally, this review seeks to shed light on the future of wearable biosensors in healthcare and wellness by summarizing existing trends and new advancements.

Developmental and Epileptic Encephalopathies: Need for Bridging the Gaps Between Clinical Syndromes and Underlying Genetic Etiologies.

Advancement in genetic testing has become increasingly important in diagnosing and managing developmental and epileptic encephalopathies (DEEs), a group of rare neurodevelopmental disorders characterized by early-onset seizures, developmental delay, and electroencephalographic (EEG) abnormalities. These early epileptic encephalopathies are often described as various syndromes as per their clearly defined, relatively uniform, and distinct clinical phenotypes with consistent EEG and/or neuroimaging findings. Finding the underlying molecular mechanisms can cause a definitive change in the management strategy. With the evolving overlapping phenotypes, advent of technologies, and discovery of new genes, it is exceedingly becoming challenging to correctly characterise these disorders and plan subsequent evidence-based management. Cytogenetic microarray (CMA) and next generation sequencing (NGS) with improved data analysis pipe-lines and algorithms have revamped the diagnostic yield dramatically. However, as of now, there is a big lacuna in step-wise evaluation guideline or consensus on integration of genetic testing results with management plan. Understanding the genetic etiologies of such syndromes timely has three major implications: (1) Knowing the outcome of such a syndrome, (2) Therapeutic implications including licensing of drugs for certain forms (e.g. genetic syndromes involving ion channels) and (3) Genetic counseling, prenatal testing and choosing reproductive options in future pregnancies in such families. The focus of this review is to provide an understanding of different types of causative variants and their step-wise genetic testing approach; the most pressing clinical need and to develop an optimal diagnostic pathway for this group of patients.

ANALYZING USER BEHAVIOR PATTERNS FOR PERSONALIZED RECOMMENDER SYSTEMS IN E-COMMERCE: A LITERATURE REVIEW

Abstract: E-commerce thrives on a user-centric strategy, and recommender systems are at the cutting edge of personalizing the purchasing experience. These systems may forecast preferences and recommend appropriate items by analyzing user behavior patterns, resulting in many benefits such as increased customer satisfaction, increased sales and conversions, and increased efficiency. To accomplish these benefits, recommender systems utilize complex algorithms that examine numerous elements of user behavior such as purchase history, browsing behavior, search queries, demographic data, and implicit feedback. Sophisticated algorithms can recognise complicated patterns in user data, resulting in more accurate and personalized suggestions. Analyzing user reviews, product descriptions, and social media interactions may help you better understand consumer preferences and product features. Systems can make real-time suggestions depending on a user’s current browsing session, resulting in a more dynamic purchasing experience. Personalized recommender systems will play an increasingly important role in molding the future of e-commerce as user behavior analysis techniques are constantly refined. The study intends to make important advances to the field of personalized recommender systems by undertaking a thorough research of user behavior patterns in the e-commerce domain. We strive to improve the performance of recommender systems by extracting insightful features from various data sources and exploring sophisticated machine learning techniques, resulting in a more engaging and tailored user experience that fosters customer satisfaction and drives business growth. A comprehensive review of user behavior patterns and their influence on personalized recommender systems in the e-commerce industry reveals the critical role of data analysis and machine learning algorithms in tailoring product suggestions to individual preferences, thereby enhancing customer satisfaction and driving sales growth. By implementing the tactics and approaches expressed in this study, e-commerce platforms may stay ahead of the curve, providing a smooth and tailored purchasing experience that surpasses customer expectations and contributes to their competitive advantage in the changing e-commerce environment.

Multiple damage detection in sandwich composite structures with lattice core using regression-based machine learning techniques

The abstract discusses the importance of Structural Health Monitoring (SHM) in ensuring the reliability and health of structures. It introduces a novel approach for recognizing cracks and delamination in sandwich composite structures using advanced methods for data analysis and machine learning algorithms. The research leverages artificial intelligence and machine learning to accurately identify and locate damage such as delamination and cracks in composite sandwich layers, which can significantly enhance troubleshooting performance, improve detection accuracy, and reduce the time and cost associated with repairs. The article presents a damage detection technique utilizing regression analysis for sandwich composite structures with a lattice core, capable of identifying and locating multiple cracks and delamination, as well as simultaneous defects in the structure. The analysis is conducted based on the sandwich composite structure’s healthy and damaged conditions in Abaqus software, and acceleration responses under random forces obtained through the finite element method were used to train various machine learning models, including regression algorithms like k-Nearest Neighborhood Regression (KNN), Light Gradient Boost Machine (LGBM), and Decision Tree Regression (DTR) to detect the damage location. The results indicate that the regression (LGBM), k-Nearest neighborhood, and decision tree regression (DTR) were the most successful functions, respectively, and the damage classification models accurately identified the damage and its location in the composite structure, achieving an accuracy rate of approximately 98.8%.

Graph Regularized Sparse L2,1 Semi‐Nonnegative Matrix Factorization for Data Reduction

ABSTRACTNon‐negative Matrix Factorization (NMF) is an effective algorithm for multivariate data analysis, including applications to feature selection, pattern recognition, and computer vision. Its variant, Semi‐Nonnegative Matrix Factorization (SNF), extends the ability of NMF to render parts‐based data representations to include mixed‐sign data. Graph Regularized SNF builds upon this paradigm by adding a graph regularization term to preserve the local geometrical structure of the data space. Despite their successes, SNF‐related algorithms to date still suffer from instability caused by the Frobenius norm due to the effects of outliers and noise. In this paper, we present a new SNF algorithm that utilizes the noise‐insensitive norm. We provide monotonic convergence analysis of the SNF algorithm. In addition, we conduct numerical experiments on three benchmark mixed‐sign datasets as well as several randomized mixed‐sign matrices to demonstrate the performance superiority of SNF over conventional SNF algorithms under the influence of Gaussian noise at different levels.

Gene signatures for cancer research: A 25-year retrospective and future avenues.

Over the past two decades, extensive studies, particularly in cancer analysis through large datasets like The Cancer Genome Atlas (TCGA), have aimed at improving patient therapies and precision medicine. However, limited overlap and inconsistencies among gene signatures across different cohorts pose challenges. The dynamic nature of the transcriptome, encompassing diverse RNA species and functional complexities at gene and isoform levels, introduces intricacies, and current gene signatures face reproducibility issues due to the unique transcriptomic landscape of each patient. In this context, discrepancies arising from diverse sequencing technologies, data analysis algorithms, and software tools further hinder consistency. While careful experimental design, analytical strategies, and standardized protocols could enhance reproducibility, future prospects lie in multiomics data integration, machine learning techniques, open science practices, and collaborative efforts. Standardized metrics, quality control measures, and advancements in single-cell RNA-seq will contribute to unbiased gene signature identification. In this perspective article, we outline some thoughts and insights addressing challenges, standardized practices, and advanced methodologies enhancing the reliability of gene signatures in disease transcriptomic research.

Retraction Note: Data analysis algorithm for internet of things based on federated learning with optical technology

Retraction Note: Data analysis algorithm for internet of things based on federated learning with optical technology

Advancing AI-Driven Linguistic Analysis: Developing and Annotating Comprehensive Arabic Dialect Corpora for Gulf Countries and Saudi Arabia

This study harnesses the linguistic diversity of Arabic dialects to create two expansive corpora from X (formerly Twitter). The Gulf Arabic Corpus (GAC-6) includes around 1.7 million tweets from six Gulf countries—Saudi Arabia, UAE, Qatar, Oman, Kuwait, and Bahrain—capturing a wide range of linguistic variations. The Saudi Dialect Corpus (SDC-5) comprises 790,000 tweets, offering in-depth insights into five major regional dialects of Saudi Arabia: Hijazi, Najdi, Southern, Northern, and Eastern, reflecting the complex linguistic landscape of the region. Both corpora are thoroughly annotated with dialect-specific seed words and geolocation data, achieving high levels of accuracy, as indicated by Cohen’s Kappa scores of 0.78 for GAC-6 and 0.90 for SDC-5. The annotation process leverages AI-driven techniques, including machine learning algorithms for automated dialect recognition and feature extraction, to enhance the granularity and precision of the data. These resources significantly contribute to the field of Arabic dialectology and facilitate the development of AI algorithms for linguistic data analysis, enhancing AI system design and efficiency. The data provided by this research are crucial for advancing AI methodologies, supporting diverse applications in the realm of next-generation AI technologies.

Assessment of the Possibility of Using the Trace Element Status of a Herbal Drug as an Additional Quality Criterion: Chamomile Flowers as a Case Study

INTRODUCTION. Herbal drugs can often be difficult to identify and distinguish from morphologically similar species, especially when fractionated. For the analysis of complex mixtures of herbal origin, it is reasonable to use additional identification methods, including hierarchical clustering of herbal drugs by their trace element content.AIM. This study aimed to assess the possibility of using hierarchical clustering by trace element content for complex biological matrices (plant mixtures), with officinal chamomile species as a case study.MATERIALS AND METHODS. Multiple morphologically similar species grow together with officinal chamomile species that are widely used in medical practice. The study examined the flowers and peduncles of Matricaria recutita L., Tanacetum parthenium (L.) Sch.Bip., Leucanthemum vulgare Lam., Tripleurospermum inodorum (L.) Sch.Bip., and Matricaria suaveolens Buchenau, as well as artificial mixtures of chamomile with adulterant species (as samples non-compliant with pharmacopoeial specifi ca tions). The content of trace elements was determined by inductively coupled plasma mass spectrometry (ICP-MS). The statistical treatment of the results used Statistica 10 software with a built-in data analysis algorithm for multivariate experiments.RESULTS. Having determined the content of 56 trace elements, the authors plotted a hierarchical cluster tree using cluster analysis methods. M. recutita and M. suaveolens were shown to be the closest to each other in terms of their trace element status, which correlated with their taxonomy. Artificial plant mixtures formed intermediate clusters that were observed between different chamomile species. Moreover, according to the criterion of dissimilarity, the trace element cluster of M. recutita significantly differed from the trace element clusters of all the adulterant species and mixtures containing more than 10% of the adulterant species.CONCLUSIONS. This study demonstrated the possibility of using hierarchical clustering based on the content of trace elements in a biological object to analyse complex systems, including plant mixtures, and identify differences that may have discriminative value for the quality control of herbal drugs.

Research on Knowledge Discovery and Sharing in AIGC Virtual Teaching and Research Room Empowered by Big Data Analysis and Natural Language Processing Algorithms

This paper introduces a pioneering framework named Deep Reinforcement Learning based AI-Generated Content for Virtual Teaching (DRL-AIGC-VR), which aims to transform the landscape of online education and research. At the heart of this system is the integration of Deep Reinforcement Learning (DRL) and Natural Language Processing (NLP), making it exceptionally suited for the dynamic and evolving environment of virtual teaching and research rooms. The uniqueness of DRLAIGC-VR lies in its adaptive content curation and presentation capabilities, achieved through a combination of Deep Q-Networks (DQN) with attention mechanisms. This innovative approach allows the system to personalize learning experiences by tailoring them to individual student performance and engagement levels. Simultaneously, it focuses on presenting the most pertinent information, thereby streamlining and optimizing the learning process. One of the most significant features of this system is its ability to handle and analyze large-scale educational data, a vital aspect in today’s big data-driven world. This capability ensures that DRL-AIGCVR offers a highly interactive, responsive, and efficient learning environment, addressing the varied requirements of students and researchers. The implementation of DRL-AIGC-VR in virtual educational settings has shown remarkable improvements in several key areas, including learning outcomes, student engagement, and knowledge retention. These enhancements are indicative of the substantial progress that the framework brings to the domain of virtual education, positioning it as a leading solution in the realm of AI-driven learning platforms. Overall, DRL-AIGC-VR represents a significant step forward in harnessing the power of AI to enrich and elevate the educational experience in virtual settings, paving the way for more advanced, personalized, and effective online learning and research methodologies.

TRIBOINFORMATICS: MACHINE LEARNING METHODS FOR FRICTIONAL INSTABILITIES

The study of friction is traditionally a data-driven area with many experimental data and phenomenological models governing structure-property relationships. Triboinformatics is a new area combining Tribology with Machine Learning (ML) and Artificial Intelligence (AI) methods, which can help to establish correlations in data on friction and wear. This is particularly relevant to unstable motion, where deterministic models are difficult to build. There are several types of friction-induced instabilities including those caused by the velocity dependency of dry friction, coupling of friction with another process (wear, heat generation, etc.), the elastic Adams instabilities, and others. The onset of sliding is also an unstable process. ML/AI methods, such as Topological Data Analysis and various ML algorithms, which have been already used for various aspects of data analysis on friction, can be applied also to the frictional instabilities.

Comparative Performance of Autoencoders and Traditional Machine Learning Algorithms in Clinical Data Analysis for Predicting Post-Staged GKRS Tumor Dynamics.

Introduction: Accurate prediction of tumor dynamics following Gamma Knife radiosurgery (GKRS) is critical for optimizing treatment strategies for patients with brain metastases (BMs). Traditional machine learning (ML) algorithms have been widely used for this purpose; however, recent advancements in deep learning, such as autoencoders, offer the potential to enhance predictive accuracy. This study aims to evaluate the efficacy of autoencoders compared to traditional ML models in predicting tumor progression or regression after GKRS. Objectives: The primary objective of this study is to assess whether integrating autoencoder-derived features into traditional ML models can improve their performance in predicting tumor dynamics three months post-GKRS in patients with brain metastases. Methods: This retrospective analysis utilized clinical data from 77 patients treated at the "Prof. Dr. Nicolae Oblu" Emergency Clinic Hospital-Iasi. Twelve variables, including socio-demographic, clinical, treatment, and radiosurgery-related factors, were considered. Tumor progression or regression within three months post-GKRS was the primary outcome, with 71 cases of regression and 6 cases of progression. Traditional ML models, such as Logistic Regression, Support Vector Machine (SVM), K-Nearest Neighbors (KNN), Extra Trees, Random Forest, and XGBoost, were trained and evaluated. The study further explored the impact of incorporating features derived from autoencoders, particularly focusing on the effect of compression in the bottleneck layer on model performance. Results: Traditional ML models achieved accuracy rates ranging from 0.91 (KNN) to 1.00 (Extra Trees). Integrating autoencoder-derived features generally enhanced model performance. Logistic Regression saw an accuracy increase from 0.91 to 0.94, and SVM improved from 0.85 to 0.96. XGBoost maintained consistent performance with an accuracy of 0.94 and an AUC of 0.98, regardless of the feature set used. These results demonstrate that hybrid models combining deep learning and traditional ML techniques can improve predictive accuracy. Conclusion: The study highlights the potential of hybrid models incorporating autoencoder-derived features to enhance the predictive accuracy and robustness of traditional ML models in forecasting tumor dynamics post-GKRS. These advancements could significantly contribute to personalized medicine, enabling more precise and individualized treatment planning based on refined predictive insights, ultimately improving patient outcomes.

Identification of two molecularly and prognostically distinct subtypes in acral melanoma using network prediction method.

Acral melanoma, characterized by its aggressiveness and poor prognosis compared to other melanoma subtypes, poses significant challenges in clinical management. However, the molecular underpinnings driving the biological and clinical features of this disease remain poorly understood. In this study, our aim was to elucidate the molecular landscape and the correlation between subtypes and clinical features of acral melanoma. We conducted comprehensive analyses to dissect the molecular characteristics of acral melanoma, employing a combination of multi-omics data analysis and network-based disease gene prediction algorithms. Single-cell RNA-Seq data were utilized to investigate the contribution of immunocytes to the molecular classification of acral melanoma. Additionally, we used clinical samples to validate the correlation between new subtypes and the prognosis of acral melanoma and the expression of subtype markers and verified the interaction between macrophages and acral melanoma cells at cellular level. Our study reveals the existence of two distinct subtypes of acral melanoma exhibiting marked differences in clinical behaviour, cellular and molecular mechanisms. We identified a robust biomarker panel (EREG, VSIG4, FCGR3A and RAB20) that accurately distinguishes these two subtypes with an impressive AUC of 0.946, validated using clinical samples. Subtype I, characterized by thinner Breslow thickness, demonstrates a favourable prognosis, whereas Subtype II represents a high-risk subtype with a propensity for dermal invasion. Notably, the signature gene EREG of Subtype I is enriched in FCN1+ macrophages, known for promoting inflammatory and immune responses. Conversely, signature genes VSIG4 and FCGR3A of Subtype II are enriched in SPP1+ macrophages, which exhibit significant crosstalk with tumour cells. Our findings significantly enhance the understanding of the molecular landscape of acral melanoma and offer novel insights into its clinical management by identifying distinct subtypes and potential therapeutic targets. The findings have to be confirmed in different cohorts in the future for full validation.