Exploration on Scientific Research Data-Targeted Intelligent Recommendation System Using Machine Learning Under the Background of Sustainable Development.

ZusammenfassungDie Nutzung von Datenbeständen für medizinische Forschung bietet großes gesamtgesellschaftliches Potenzial, können sich doch durch die Auswertung großer Datenmengen mit maschinellen Lernverfahren teils ganz neue Forschungsansätze, Diagnose- und Behandlungsmethoden ergeben. Häufig scheitert eine solche Nutzung von Datenbeständen aber an den hohen Voraussetzungen oder unklaren Anforderungen des Datenschutzrechts.Gerade die Verarbeitung visueller Gesundheitsdaten wie MRT-Gehirnscans geht mit besonderen Risiken für die betroffenen Personen einher, die die rechtmäßige Verarbeitung dieser Daten zu Forschungszwecken erschweren. Zur Reduzierung dieser Risiken können als unabhängige Anonymisierungs- und Pseudonymisierungsinstanz eingesetzte Datentreuhänder beitragen, die eine sichere Verarbeitungsumgebung bereitstellen, in der Gesundheitsdaten kurzfristig gespeichert, ausgewertet und anschließend gelöscht werden.Diese Risikoreduktion trägt zur Rechtmäßigkeit der Datenverarbeitung bei, führt sie doch dazu, dass Abwägungsentscheidungen, wie z. B. nach Art. 9 Abs. 2 lit. j, Art. 89 Abs. 1 Datenschutzgrundverordnung (DSGVO), § 6 Abs. 2 Gesundheitsdatenschutzgesetz Nordrhein-Westfalen (GDSG NW), § 17 Datenschutzgesetz Nordrhein-Westfalen (DSG NRW) bzw. § 27 Bundesdatenschutzgesetz (BDSG), oder die Beurteilung der Kompatibilität des Sekundärzwecks der Verarbeitung mit dem Erhebungszweck eher zugunsten der Datenverarbeitung ausfällt. Der Einsatz von Datentreuhandmodellen kann insofern auch zum Abbau datenschutzrechtlicher Hürden der Verarbeitung solcher Gesundheitsdaten zu wissenschaftlichen Forschungszwecken beitragen.

One of the major challenges faced in oil and gas industry today is talent management of its workforce. Training the workforce and suggesting the right training courses to the individuals is important in talent management and career development. There are many training recommendation systems available using different machine learning approaches such as collaborative filtering, neural networks and hybrid models. In this paper, an intelligent recommendation system is proposed by blending machine learning algorithms, natural language processing (NLP) and text analytics combined with organizational preferences. This framework gives a recommender system considering the user profiles, training preferences for his or her organization where each set of organizational units will have unique training recommendation requirements considering organizational functional behavior. The proposed mechanism uses machine learning algorithms at different stages of its learning process and ensemble them in a unique fashion such that desirable results are achieved to the user satisfaction.

The explosion of internet news data has led to frequent information overload, which in turn has caused the dual problems of low recommendation quality and poor user experience. The research combines machine learning with image matching algorithms to provide users with interesting news content. Firstly, the news recommendation algorithm based on multi-task learning (MTL) constructs the input sequence through the users click history, maps the inputs of different tasks to the shared space and extracts text features. The image matching algorithm is adopted to record users historical preferences and capture the changing trends of interests. Experimental data show that the accuracy rate of the content-based recommendation (CB) algorithm reaches 78.27%. The collaborative filtering (CF) recommendation algorithm reached 87.97%. The recommendation accuracy of the joint recommendation algorithm all exceeds 90%. Moreover, the accuracy of the joint recommendation algorithm significantly exceeds that of the two traditional recommendation methods, CB and CF.

In the ever-changing realm of E-Commerce, it is essential for online businesses to comprehend and adjust to shifting consumer behaviour in order to achieve long-term success. In which, Intelligent Recommendation System (IRS) has gained familiarity by suggesting personalized information based on user preference and behaviours. Hence, the review paper primarily aims to analyse significance of the intelligent recommendation system to transform ecommerce field, specifically enrich the user personalisation and satisfaction, and enhance revenue in business. Accordingly, the proposed survey is discussed the traditional system and AI-powered personalization system in ecommerce. AI-powered recommendation system utilize sophisticated algorithms to analyse extensive data, allowing for the provision of highly customized and relevant content, product recommendation, and user satisfaction. Besides, it examines future trends in AI integration within e-commerce, particularly advancements in Natural Language Processing (NLP) and visual search technologies, which are poised to further enrich ecommerce. The paper concludes with a look toward future directions for the integration of AI technologies in e-commerce, anticipating advancements in NLP and visual search capabilities, which promise to further enhance the online shopping experience. Overall, the findings of the article underscores the transformative impact of IRS on the e-commerce sector, advocating for their continued development in response to evolving market demands.

Traditional recommendation algorithms face some serious problems, including data sparsity, cold start and inefficiency. To better address the problems above, the paper proposes a hybrid recommendation algorithm based on improved collaborative filtering of user context fuzzy clustering and content-based. For collaborative filtering, firstly, user classification is based on fuzzy clustering according to user context, and then collaborative filtering is used to recommend products for similar users. And the improved content-based algorithm sets up feature vectors for users and items dynamically. Experiments show that the hybrid algorithm can avoid defects of single algorithm and improve the performance in both recommendation quality and efficiency, which opens up exciting avenues for future research.

Neuroimaging MRI data in scientific research is increasingly pooled, but the reliability of such studies may be hampered by the use of different hardware elements. This might introduce bias, for example when cross-sectional studies pool data acquired with different head coils, or when longitudinal clinical studies change head coils halfway. In the present study, we aimed to estimate this possible bias introduced by using different head coils to create awareness and to avoid misinterpretation of results. We acquired, with both an 8 channel and 32 channel head coil, T1-weighted, diffusion tensor imaging and resting state fMRI images at 3T MRI (Philips Achieva) with stable acquisition parameters in a large group of cognitively healthy participants (n = 77). Standard analysis methods, i.e., voxel-based morphometry, tract-based spatial statistics and resting state functional network analyses, were used in a within-subject design to compare 8 and 32 channel head coil data. Signal-to-noise ratios (SNR) for both head coils showed similar ranges, although the 32 channel SNR profile was more homogeneous. Our data demonstrates specific patterns of gray and white matter volume differences between head coils (relative volume change of 6 to 9%), related to altered image contrast and therefore, altered tissue segmentation. White matter connectivity (fractional anisotropy and diffusivity measures) showed hemispherical dependent differences between head coils (relative connectivity change of 4 to 6%), and functional connectivity in resting state networks was higher using the 32 channel head coil in posterior cortical areas (relative change up to 27.5%). This study shows that, even when acquisition protocols are harmonized, the results of standardized analysis models can be severely affected by the use of different head coils. Researchers should be aware of this when combining multiple neuroimaging MRI datasets, to prevent coil-related bias and avoid misinterpretation of their findings.

The estimation of retrieval uncertainty and stability are essential for the accurate interpretation of data in scientific research, use in analyses, or numerical models. The primary uncertainty sources of satellite SST retrievals are due to errors in spacecraft navigation, sensor calibration, sensor noise, retrieval algorithms, and incomplete identification of corrupted retrievals. In this study, comparisons to in situ data are utilized to investigate retrieval accuracies of microwave (MW) SSTs from the Advanced Microwave Scanning Radiometer—Earth Observing System (AMSR-E) and infrared (IR) SSTs from the Moderate Resolution Imaging Spectroradiometer (MODIS). The highest quality MODIS data were averaged to 25 km for comparison. The in situ SSTs are used to determine dependencies on environmental parameters, evaluate the identification of erroneous retrievals, and examine biases and standard deviations (STD) for each of the satellite SST data sets. Errors were identified in both the MW and IR SST data sets: (1) at low atmospheric water vapor a posthoc correction added to AMSR-E was incorrectly applied and (2) there is significant cloud contamination of nighttime MODIS retrievals at SST <10°C. A correction is suggested for AMSR-E SSTs that will remove the vapor dependency. For MODIS, once the cloud contaminated data were excluded, errors were reduced but not eliminated. Biases were found to be −0.05°C and −0.13°C and standard deviations to be 0.48°C and 0.58°C for AMSR-E and MODIS, respectively. Using a three-way error analysis, individual standard deviations were determined to be 0.20°C (in situ), 0.28°C (AMSR-E), and 0.38°C (MODIS).

This study explores the main convergences and singularities of data protection and privacy laws in South America. It begins with a historical overview of privacy rights, tracing their evolution from the foundational concept of the “right to be left alone” to the contemporary legal emphasis on consent, personal data lifecycle management, and information security. The research employs a qualitative methodology combining content analysis and the study of foreign law and using the ATLAS.ti software, the authors coded and analyzed privacy and data protection legislation from eight South American countries: Argentina, Brazil, Chile, Colombia, Ecuador, Paraguay, Peru, and Uruguay. The findings reveal a significant degree of alignment among these laws, not only in common terminology such as data subject, controller, and data processing, but also in a shared normative core of principles and data subject rights. Despite these similarities, the study highlights key national distinctions. Brazil’s legislation uniquely defines the use of personal data in scientific research. Ecuadorian law stands out for recognizing the rights of deceased individuals and establishing certifying entities to oversee data protection compliance. In contrast, Paraguay's law is explicitly focused on personal credit data, reflecting a more limited scope. By comparing these regional legal frameworks, the article contributes to a deeper understanding of how global privacy standards are adapted within diverse national contexts in South America.

It has been argued that the use of big data in scientific research obviates the need for causal knowledge in making sound predictions and interventions. Whilst few accept that this claim is true, there is an ongoing discussion about what effect, if any, big data has on scientific methodology and, in particular, the search for causes. One response has been to show that the automated analysis of big data by a computer program can be used to find causes in addition to mere correlations. However, up until now it has only been demonstrated how this can be achieved with respect to difference-making causes. Yet it is widely acknowledged that scientists need evidence of both “difference-making” and “production” in order to infer a genuine causal link. This paper fills in the gap by outlining how computer-assisted discovery in big data can find productive causes. This is achieved by developing an inference rule based on a little-known causal process theory called the information transmission account.

Crowdsourcing: A Critical Reflection on This New Frontier of Participant Recruiting in Nutrition and Dietetics Research

Research on the Construction of Statistical Data Management Platform under Big Data

Due to the low cognitive ability of children, the intelligent recommendation system should assist children in learning with the assistance of parents. Therefore, the intelligent recommendation system needs to evaluate the ability of users (children learners and their parents), and assign learning content with suitable difficulty Match with the user’s ability, assist the user to learn efficiently, and gradually cultivate the user’s learning interest and habits, and can be a useful supplement to the user’s regular learning. This paper uses JAVA EE language to develop an intelligent recommendation system for children’s learning platform based on Hadoop big data framework and machine learning algorithms. Learning platform with intelligent recommendation function. By analyzing the cognitive ability of children, and using the collaborative recommendation algorithm to recommend learning resources suitable for this age stage for children, in order to improve learning efficiency and cultivate children’s learning interest from an early age, at the same time, the resource library in the learning platform has more Practical value.

Intelligent recommender systems are essential for identifying at-risk students and personalizing learning through tailored resources. Accurate prediction of student performance enables these systems to deliver timely interventions and data-driven support. This paper presents the application of machine learning models to predict final exam grades in a university-level programming course, leveraging multi-modal student data to improve prediction accuracy. In particular, a recent raw dataset of students enrolled in a programming course across 36 class sections from the Fall 2024 and Winter 2025 terms was initially processed. The data was collected up to one month before the final exam. From this data, a comprehensive set of features was engineered, including the student’s background, assessment grades and completion times, digital learning interactions, and engagement metrics. Building on this feature set, six machine learning prediction models were initially developed using data from the Fall 2024 term. Both training and testing were conducted on this dataset using cross-validation combined with hyperparameter tuning. The XGBoost model demonstrated strong performance, achieving an accuracy exceeding 91%. To assess the generalizability of the considered models, all models were retrained on the complete Fall 2024 dataset. They were then evaluated on an independent dataset from Winter 2025, with XGBoost achieving the highest accuracy, exceeding 84%. Feature importance analysis has revealed that the midterm grade and the average completion duration of lab assessments are the most influential predictors. This data-driven approach empowers instructors to proactively identify and support at-risk students, enabling adaptive learning environments that deliver personalized learning and timely interventions.

Aim: The majority of machine learning (ML) models in healthcare are built on retrospective data, much of which is collected without explicit patient consent for use in artificial intelligence (AI) and ML applications. The primary aim of this study was to evaluate whether clinicians and scientific researchers themselves consent to provide their own data for the training of ML models. Materials and Methods: This survey was conducted through an anonymous online survey, utilizing platforms such as Telegram, LinkedIn, and Viber. The target audience comprised specific international groups, primarily Russian, German, and English-speaking, of clinicians and scientific researchers. These participants ranged in their levels of expertise and experience, from beginners to veterans. The survey centered on a singular, pivotal question: “Do You Consent to the Use of Your Biological and Private Data for Training Machine Learning and AI Models?” Respondents had the option to choose from three responses: “Yes” and “No”. Results: The survey was conducted in January 2024. A total of 119 unique and verified individuals participated in the survey. The results revealed that only 50% of respondents (63 persons) expressed consent to provide their own data for the training of ML and AI models. Conclusion: In the development of ML and AI models, particularly open-source ones, it is crucial to ascertain whether participants are willing to provide their private data. While ML algorithms can transform the nature of data, it is important to remember that the primary owner of this data is the individual. Our findings show that in 50% of the cases, even participants from scientific research and clinical backgrounds – individuals typically accountable for ensuring data quality in AI and ML model development – do not consent to the use of their data in AI and ML settings. This highlights the need for more stringent consent processes and ethical considerations in the utilization of personal data in AI and ML research.

In this paper, we present an FDO[1] implementation focusing on standardized scientific research data exchange. Using DOIP[2] and Cordra[3] and based on a concrete FDO model with assumed layers and compositions, we implement some basic use-cases of exchanging/handling scientific research data based on FDOs, and introduce FDO services to handle research data based on FDOs. We introduce workflows to orchestrate the application of FDO services. These define and foster an FDO-based ecosystem research data. Our work was funded by the German Research Foundation(DFG) within the KonsortSWD, the consortium for the social, behavioral, educational and economic sciences, is part of the German National Research Data Infrastructure (NFDI) initiative.