Data Science Methods Research Articles

Probing the randomness of the local current distributions of 316 L stainless steel corrosion in NaCl solution

This investigation proposes using Scanning Electrochemical Cell Microscopy (SECCM) as a high throughput tool to collect corrosion activity from randomly probed locations on 316 L SS. In the presence of chloride, potentiodynamic polarisation tests triggered the development of pitting corrosion. Data science methods were deployed to handle and explore 955 j Vs E curves. Normality tests and fitting with theoretical functions were used to understand the conditional log(j) distributions at different potentials. Unimodal and uniform distributions were assigned to the passive and pitting regions. Our “big-data” local strategy revealed a potential-dependent distribution of log(j), with the randomness increasing with testing aggressiveness. Data availabilityAll data generated or analysed during this study are included in this published article (and its supplementary information files) and are available in the Mendeley Data repository, [https://data.mendeley.com/datasets/78rz8vw46x/2]. Code availabilityThe code required to reproduce these findings is included in this published article (and its supplementary information files) and is available to download from GitHub: https://github.com/bcoelho-leonardo/Data-driven-analysis-of-the-local-current-distributions-of-316L-corrosion-in-NaCl-solution/blob/4efff485b115468840b25ea56ad81b31711c0f51/local%20current%20distributions%20of%20316L%20corrosion.ipynb.

An approach for analyzing urban carbon emissions using machine learning models

Carbon peaking and carbon neutrality goals have posed great challenges to transforming local economies into low-carbon economies. Hence, establishing an effective carbon management system is urgent. However, the development of the urban carbon management system is hampered by the immaturity of the carbon emission accounting system at the city level. To compensate for the insufficiency of the existing urban carbon emission accounting system and to find the city government in constructing a perfect carbon emission management system as soon as possible, this study used the data science method based on the statistical data of 285 cities in China from 2005 to 2017 to explore the influencing factors of urban carbon emissions, that is, using light gradient boosting machine and the accumulated local effects interpretable models to screen potential influencing factors of urban carbon emissions. Then, an index system for urban carbon management was evaluated and proposed, and a case analysis was conducted with urban industrial electricity consumption as a background. This method can be easily integrated with the existing urban management system, which could reduce the cost of building a carbon management system.

What is mineral informatics?

Abstract Minerals are information-rich materials that offer researchers a glimpse into the evolution of planetary bodies. Thus, it is important to extract, analyze, and interpret this abundance of information to improve our understanding of the planetary bodies in our solar system and the role our planet’s geosphere played in the origin and evolution of life. Over the past several decades, data-driven efforts in mineralogy have seen a gradual increase. The development and application of data science and analytics methods to mineralogy, while extremely promising, has also been somewhat ad hoc in nature. To systematize and synthesize the direction of these efforts, we introduce the concept of “Mineral Informatics,” which is the next frontier for researchers working with mineral data. In this paper, we present our vision for Mineral Informatics and the X-Informatics underpinnings that led to its conception, as well as the needs, challenges, opportunities, and future directions of the field. The intention of this paper is not to create a new specific field or a sub-field as a separate silo, but to document the needs of researchers studying minerals in various contexts and fields of study, to demonstrate how the systemization and enhanced access to mineralogical data will increase cross- and interdisciplinary studies, and how data science and informatics methods are a key next step in integrative mineralogical studies.

Can NeuroIS improve executive employee recruitment? Classifying levels of executive functions using resting state EEG and data science methods

Based on the assumption that the success of an organization is largely determined by the knowledge and skills of its employees, human resource (HR) departments invest considerable resources in the employee recruitment process with the aim of selecting the best, most suitable employees. Due to the high cost of the recruitment process along with its high rate of uncertainty, HR recruiters utilize a variety of methods and instruments to improve the efficiency and effectiveness of this process. Thus far, however, neurological methods, in which neurobiological signals from an examined person are analyzed, have not been utilized for this purpose. This study is the first to propose a neuro-based decision support system to classify cognitive functions into levels, whose target is to enrich the information and indications regarding the candidate along the employee recruitment processes. We first measured relevant functional and cognitive abilities of 142 adult participants using traditional computer-based assessment, which included a battery of four tests regarding executive functions and intelligence score, consistent with actual recruitment processes. Second, using electroencephalogram (EEG) technology, which is one of the dominant measurement tools in NeuroIS research, we collected the participants' brain signals by administering a resting state EEG (rsEEG) on each participant. Finally, using advanced machine and deep learning algorithms, we leveraged the collected rsEEG to classify participants' levels of executive functions and intelligence score. Our empirical analyses show encouraging results of up to 72.6% accuracy for the executive functions and up to 71.2% accuracy for the intelligence score. Therefore, this study lays the groundwork for a novel, generic (non-stimuli based) system that supports the current employee recruitment processes, that is based on psychological theories of assessing executive functions. The proposed decision support system could contribute to the development of additional medium of assessing employees remotely which is especially relevant in the current Covid-19 pandemic. While our method aims at classification rather than at explanation, our intriguing findings have the potential to push forward NeuroIS research and practice.

Effect of a Machine Learning Recommender System and Viral Peer Marketing Intervention on Smoking Cessation

Novel data science and marketing methods of smoking-cessation intervention have not been adequately evaluated. To compare machine learning recommender (ML recommender) computer tailoring of motivational text messages vs a standard motivational text-based intervention (standard messaging) and a viral peer-recruitment tool kit (viral tool kit) for recruiting friends and family vs no tool kit in a smoking-cessation intervention. This 2 ×2 factorial randomized clinical trial with partial allocation, conducted between July 2017 and September 2019 within an online tobacco intervention, recruited current smokers aged 18 years and older who spoke English from the US via the internet and peer referral. Data were analyzed from March through May 2022. Participants registering for the online intervention were randomly assigned to the ML recommender or standard messaging groups followed by partially random allocation to access to viral tool kit or no viral tool kit groups. The ML recommender provided ongoing refinement of message selection based on user feedback and comparison with a growing database of other users, while the standard system selected messages based on participant baseline readiness to quit. Our primary outcome was self-reported 7-day point prevalence smoking cessation at 6 months. Of 1487 participants who smoked (444 aged 19-34 years [29.9%], 508 aged 35-54 years [34.1%], 535 aged ≥55 years [36.0%]; 1101 [74.0%] females; 189 Black [12.7%] and 1101 White [78.5%]; 106 Hispanic [7.1%]), 741 individuals were randomly assigned to the ML recommender group and 746 individuals to the standard messaging group; viral tool kit access was provided to 745 participants, and 742 participants received no such access. There was no significant difference in 6-month smoking cessation between ML recommender (146 of 412 participants [35.4%] with outcome data) and standard messaging (156 of 389 participants [40.1%] with outcome data) groups (adjusted odds ratio, 0.81; 95% CI, 0.61-1.08). Smoking cessation was significantly higher in viral tool kit (177 of 395 participants [44.8%] with outcome data) vs no viral tool kit (125 of 406 participants [30.8%] with outcome data) groups (adjusted odds ratio, 1.48; 95% CI, 1.11-1.98). In this study, machine learning-based selection did not improve performance compared with standard message selection, while viral marketing did improve cessation outcomes. These results suggest that in addition to increasing dissemination, viral recruitment may have important implications for improving effectiveness of smoking-cessation interventions. ClinicalTrials.gov Identifier: NCT03224520.

Technology Platforms and Approaches for Building and Evaluating Machine Learning Methods in Healthcare.

Artificial intelligence (AI) methods are becoming increasingly commonly implemented in healthcare as decision support, business intelligence tools, or, in some cases, Food and Drug Administration-approved clinical decision-makers. Advanced lab-based diagnostic tools are increasingly becoming AI driven. The path from data to machine learning methods is an active area for research and quality improvement, and there are few established best practices. With data being generated at an unprecedented rate, there is a need for processes that enable data science investigation that protect patient privacy and minimize other business risks. New approaches for data sharing are being utilized that lower these risks. In this short review, clinical and translational AI governance is introduced along with approaches for securely building, sharing, and validating accurate and fair models. This is a constantly evolving field, and there is much interest in collecting data using standards, sharing data, building new models, evaluating models, sharing models, and, of course, implementing models into practice. AI is an active area of research and development broadly for healthcare and laboratory testing. Robust data governance and machine learning methodological governance are required. New approaches for data sharing are enabling the development of models and their evaluation. Evaluation of methods is difficult, particularly when the evaluation is performed by the team developing the method, and should ideally be prospective. New technologies have enabled standardization of platforms for moving analytics and data science methods.

APPLICATION OF DATA SCIENCE METHODS FOR DEMAND FORECASTING IN RETAIL

This scientific article examines the problem of forecasting demand in retail using data science methods. It is explained that traditional methods of demand forecasting do not give an excellent result, as machine learning, statistical models and data analysis become powerful tools, they need improvement, therefore this research is necessary and appropriate. The importance of accurate demand forecasting for effective inventory management, cost reduction, and customer service improvement is analyzed. The main methods of data science are considered, such as: machine learning, statistical models and data analysis. Real examples of the use of these methods in retail companies and their impact on increasing the accuracy of demand forecasting are also presented and clearly characterized for each company. Key steps in the forecasting process are described, including data collection and preparation, model selection, training, and performance evaluation. Examples of the use of machine learning algorithms, such as linear regression, decision trees, and neural networks, for demand forecasting in the retail sector are provided, and a comparison of these approaches is highlighted. The proposed price optimization procedure. This article shows that forecasting and analytics are integral to the effectiveness and competitiveness and flexibility of retailers in the market, and that the results of this study can be widely applied to further study the application of these methods, as well as to identify new methods. According to the scientific article, a conclusion was made that this research should be continued, and it will contribute to the effective functioning of retail companies and improve their competitiveness on the market. Recent achievements and prospects of using data science in demand forecasting are discussed.

Inaugural Editorial. Can We Achieve Our Mission: Fast, Accessible, Cutting-edge, and Top-quality?

We are pleased to launch the first issue of the New England Journal of Statistics in Data Science (NEJSDS). NEJSDS is the official journal of the New England Statistical Society (NESS) under the leadership of Vice President for Journal and Publication and sponsored by the College of Liberal Arts and Sciences, University of Connecticut. The aims of the journal are to serve as an interface between statistics and other disciplines in data science, to encourage researchers to exchange innovative ideas, and to promote data science methods to the general scientific community. The journal publishes high quality original research, novel applications, and timely review articles in all aspects of data science, including all areas of statistical methodology, methods of machine learning, and artificial intelligence, novel algorithms, computational methods, data management and manipulation, applications of data science methods, among others. We encourage authors to submit collaborative work driven by real life problems posed by researchers, administrators, educators, or other stakeholders, and which require original and innovative solutions from data scientists.

Evaluating probabilistic forecasts for maritime engineering operations

Abstract Maritime engineering relies on model forecasts for many different processes, including meteorological and oceanographic forcings, structural responses, and energy demands. Understanding the performance and evaluation of such forecasting models is crucial in instilling reliability in maritime operations. Evaluation metrics that assess the point accuracy of the forecast (such as root-mean-squared error) are commonplace, but with the increased uptake of probabilistic forecasting methods such evaluation metrics may not consider the full forecasting distribution. The statistical theory of proper scoring rules provides a framework in which to score and compare competing probabilistic forecasts, but it is seldom appealed to in applications. This translational paper presents the underlying theory and principles of proper scoring rules, develops a simple panel of rules that may be used to robustly evaluate the performance of competing probabilistic forecasts, and demonstrates this with an application to forecasting surface winds at an asset on Australia’s North–West Shelf. Where appropriate, we relate the statistical theory to common requirements by maritime engineering industry. The case study is from a body of work that was undertaken to quantify the value resulting from an operational forecasting product and is a clear demonstration of the downstream impacts that statistical and data science methods can have in maritime engineering operations.

A systematic study of transfer learning for colorectal cancer detection

Background and objectiveWith the rapid development of data science methods like deep learning, these methods have already been used into the field of healthcare and medicine. However, due to regulations and ethical issues, it is practically difficult to obtain large amount of medical data for training deep learning models. Transfer learning is a powerful tool to reuse the knowledge gained from different domains, which makes it possible to retrain deep learning models only with small datasets in medical image processing. In this contribution, a systematic study of model transfer techniques like fine tuning parts of the network or adding additional layers, for medical image data was conducted. MethodsThe study accomplished a binary classification task based on a colorectal cancer dataset, including microsatellite unstable or hypermutated (MSIMUT) and microsatellite stable (MSS) images. By using K-fold cross-validation, the performances of five pretrained models (DenseNet121, DenseNet201, InceptionV3, MobileNetV2 and ResNet50) were assessed according to balanced accuracy. As baseline methods, combinations of transfer learning as feature extractor and principal component analysis with linear discriminant analysis (PCA-LDA) or support vector machine (PCA-SVM) were utilised, and compared with the transfer learning counterparts. ConclusionsThe results have shown that adding convolutional layers perform obviously better than simply using the original network or fine-tuning some last layers of the network. Furthermore, a proposed bagging method performed well on a testing dataset. This study reduces the workload for future transfer learning tasks in the biomedical domain and allows to test promising transfer learning strategies first.

DESCRIBING THE EVOLUTION OF MEDICATION USE OVER TIME IN PEOPLE LIVING WITH DEMENTIA USING NETWORK ANALYSIS

Abstract Prescribing for community-dwelling older adults living with dementia is complex. Multiple medications may be used to manage symptoms associated with dementia and/or co-existing chronic conditions, and can lead to problematic polypharmacy. Our objective was to use network analysis, a data science method, to provide a comprehensive description of co-prescribed medications in persons with dementia and describe whether these patterns change over time. We created a population-based cohort of community-dwelling older adults (aged 67+ years) in Ontario, Canada, newly diagnosed with dementia (between April 2014 and January 2019), from health administrative data, and developed medication networks at one year prior to, at, and for up to five years following dementia diagnosis. Among 136,292 individuals newly diagnosed with dementia, the mean age was 82.2 years and 59% were female. The most common medication subclasses dispensed at diagnosis were primarily cardiovascular medications: statins (45.6%), proton pump inhibitors (27.3%), beta-blockers (27.0%), calcium blockers (25.1%), and ACE inhibitors (24.6%). Similar proportions of medication subclasses were found at five years after diagnosis, except cholinesterase inhibitors (34.0% at five years were dispensed cholinesterase inhibitors compared to 16.9% at diagnosis). The most frequent co-prescribed medication pairs at diagnosis included statins and beta-blockers (16.0%), proton pump inhibitors (16.0%), and ace inhibitors (15.4%), respectively. Co-prescription was similar at five years, but also included higher frequency of co-prescribing with cholinesterase inhibitors (e.g., 19.4% were prescribed cholinesterase inhibitors and statins). Network diagrams demonstrate the complexity of prescribing in this population and highlight concurrent prescribing which may require careful monitoring or deprescribing.

Thermal design and analysis of a passive modular molten salt microreactor concept

Countries and organizations around the world are in need of better energy sources that meet modern increases of demand and flexibility, such as remote power capabilities. Nuclear fission reactors have the potential to meet these needs but are not viable in their current form, due to four major challenges: proliferation concerns, economics, waste management, and a lack of passive accident prevention. Many entities in different countries are developing Generation IV reactors, particularly molten salt reactors, to try to overcome these challenges and meet their countries’ needs. However, all current designs fall short of optimal accident prevention, and have various challenges in the area of proliferation, economics, and waste management. The molten salt microreactor (MSMR) is a concept created by researchers at Brigham Young University designed to overcome each of the major challenges of current nuclear power with a uniquely simple concept. This paper presents a proof of concept for the MSMR, and lays the foundation for future design optimization. Among other innovations, the MSMR design strategically employs thermally conductive plates to help remove heat during unplanned transients. The modeling was completed in STARCCM+ and confirms the MSMR’s capability to remain below safe operating limits (1204 °C for Hastelloy, 1570 °C for FLiNaK, and 1084 °C for copper) for steady state operation and during key transients. The result of this analysis is a robust design that appears promising given the initial scoping analyses conducted in this work. Future work will involve fully optimizing the MSMR’s design in terms of coolant tube location, fin size and frequency, plate thickness, reactor percentage modeled, and material selection for structures and fuel. These modifications could potentially improve the reactor’s temperature profiles, fuel lifetime, and material lifetime. The MSMR design and simulation data will also be used in conjunction with data science methods to develop beneficial correlations and design procedures to greatly increase the speed of future reactor design for any reactor concept.

Supporting responsible machine learning in heliophysics

Over the last decade, Heliophysics researchers have increasingly adopted a variety of machine learning methods such as artificial neural networks, decision trees, and clustering algorithms into their workflow. Adoption of these advanced data science methods had quickly outpaced institutional response, but many professional organizations such as the European Commission, the National Aeronautics and Space Administration (NASA), and the American Geophysical Union have now issued (or will soon issue) standards for artificial intelligence and machine learning that will impact scientific research. These standards add further (necessary) burdens on the individual researcher who must now prepare the public release of data and code in addition to traditional paper writing. Support for these is not reflected in the current state of institutional support, community practices, or governance systems. We examine here some of these principles and how our institutions and community can promote their successful adoption within the Heliophysics discipline.

The triangulation of ethical leader signals using qualitative, experimental, and data science methods

To advance ethical leadership using signaling theory, the current work presents a mixture of inductive and deductive studies. Using a constant comparative analysis method, Study 1 involved coding CEO letters to shareholders (n = 10,919 sentences). Eight verbal ethical leader signals (ELSs) emerged and were associated with emotions (e.g., righteous anger, pride). In a set of preregistered experiments, ELSs were found to lead to evaluations of ethical leadership (Study 2: n = 264; Cohen’s d = 0.26). Study 3 illustrated that ELSs led to a reduction in financial theft (n = 434; Cohen’s d = 0.20). Study 4 showed that ELSs led to an improvement in performance (n = 434; Cohen’s d = 0.18) but had little effect on extra role behavior (Cohen’s d = 0.06). Finally, in Study 5 a machine learning algorithm, DeepEthics, was created to automatically score text (ROC =. 84; r = 0.85 between human and algorithm scores), such as emails and meeting transcripts, for ELSs in future research. Recommendations for theory and practice are discussed.

Theoretical Virtues in Scientific Practice: An Empirical Study

It is a common view among philosophers of science that theoretical virtues (also known as epistemic or cognitive values), such as simplicity and consistency, play an important role in scientific practice. In this article, I set out to study the role that theoretical virtues play in scientific practice empirically. I apply the methods of data science, such as text mining and corpus analysis, to study large corpora of scientific texts in order to uncover patterns of usage. These patterns of usage, in turn, might shed some light on the role that theoretical virtues play in scientific practice. Overall, the results of this empirical study suggest that scientists invoke theoretical virtues explicitly, albeit rather infrequently, when they talk about models (less than 30%), theories (less than 20%), and hypotheses (less than 15%) in their published works. To the extent that they are mentioned in scientific publications, the results of this study suggest that accuracy, consistency, and simplicity are the theoretical virtues that scientists invoke more frequently than the other theoretical virtues tested in this study. Interestingly, however, depending on whether they talk about hypotheses, theories, or models, scientists may invoke one of those theoretical virtues more than the others.

USAGE OF IT TECHNOLOGIES IN MEDICINE AND GENOMICS

In this article, we will consider what IT technologies are most used in medicine and by genomics methods in particular, also we will take a look at the use of big data in this matter. Additionally, we will learn what a connectome is, analyze 4M and 3V frameworks in genomics. Statistics in medicine is one of the analysis tools experimental data and clinical observations, as well as the language by means of which the obtained mathematical results are reported. However, this is not the only task of statistics in medicine. Mathematical apparatus widely used for diagnostic purposes, solving classification problems and search for new patterns, for setting new scientific hypotheses. The use of statistical programs presupposes knowledge of the basic methods and stages of statistical analysis: their sequence, necessity and sufficiency. In the proposed presentation, the main emphasis is not on detailed presentation of the formulas that make up the statistical methods, and on their essence and application rules. Finally, we talk through genome-wide association studies, methods of statistical processing of medical data and their relevance. In this article, we analyzed the basic concepts of statistics, statistical methods in medicine and data science, considered several areas in which large amounts of data are used that require modern IT technologies, including genomics, genome-wide association studies, visualization and connectome data collection.

Mental health progress requires causal diagnostic nosology and scalable causal discovery.

Nine hundred and seventy million individuals across the globe are estimated to carry the burden of a mental disorder. Limited progress has been achieved in alleviating this burden over decades of effort, compared to progress achieved for many other medical disorders. Progress on outcome improvement for all medical disorders, including mental disorders, requires research capable of discovering causality at sufficient scale and speed, and a diagnostic nosology capable of encoding the causal knowledge that is discovered. Accordingly, the field's guiding paradigm limits progress by maintaining: (a) a diagnostic nosology (DSM-5) with a profound lack of causality; (b) a misalignment between mental health etiologic research and nosology; (c) an over-reliance on clinical trials beyond their capabilities; and (d) a limited adoption of newer methods capable of discovering the complex etiology of mental disorders. We detail feasible directions forward, to achieve greater levels of progress on improving outcomes for mental disorders, by: (a) the discovery of knowledge on the complex etiology of mental disorders with application of Causal Data Science methods; and (b) the encoding of the etiological knowledge that is discovered within a causal diagnostic system for mental disorders.

Identifying the determinants of lapse rates in life insurance: an automated Lasso approach

Lapse risk is a key risk driver for life and pensions business with a material impact on the cash flow profile and the profitability. The application of data science methods can replace the largely manual and time-consuming process of estimating a lapse model that reflects various contract characteristics and provides best estimate lapse rates, as needed for Solvency II valuations. In this paper, we use the Lasso method which is based on a multivariate model and can identify patterns in the data set automatically. To identify hidden structures within covariates, we adapt and combine recently developed extended versions of the Lasso that apply different sub-penalties for individual covariates. In contrast to random forests or neural networks, the predictions of our lapse model remain fully explainable, and the coefficients can be used to interpret the lapse rate on an individual contract level. The advantages of the method are illustrated based on data from a European life insurer operating in four countries. We show how structures can be identified efficiently and fed into a highly competitive, automatically calibrated lapse model.

Enhancing predictive skills in physically-consistent way: Physics Informed Machine Learning for hydrological processes

Current modeling approaches in hydrology often rely on either physics-based or data-science methods, including Machine Learning (ML) algorithms. While physics-based models tend to have rigid structures resulting in unrealistic parameter values in certain instances, ML algorithms establish the input–output relationship while ignoring the constraints imposed by well-known physical processes. While there is a notion that the physics-based model enables better process understanding and ML algorithms exhibit better predictive skills, scientific knowledge that does not add to predictive ability may be deceptive. Hence, there is a need for a hybrid modeling approach to couple ML algorithms and physics-based models in a synergistic manner. Here we develop the Physics Informed Machine Learning (PIML) model that combines the process understanding of conceptual hydrological model with predictive abilities of state-of-the-art ML models. We apply the proposed model to predict the monthly time series of the target (streamflow) and intermediate variables (actual evapotranspiration) in ten different subcatchments in peninsular India. Our results show the capability of the PIML model to outperform a purely conceptual model (abcd model) and ML algorithms while ensuring the physical consistency in outputs validated through water balance analysis. The systematic approach for combining conceptual model structure with ML algorithms could be used to improve the predictive accuracy of crucial hydrological processes important for flood risk assessment.

Strengthening capacities of multidisciplinary professionals to apply data science in public health: Experience of an international graduate diploma program in Peru

Nowadays it is necessary to strengthen health information systems and data-based solutions. However, there are few graduate training programs in Peru to use tools and methods of data science applied in public health. This article describes the development process and the initial assessment regarding the experience of the participants in an international multidisciplinary diploma in data intelligence for pandemics and epidemics preparedness, which was carried out from January to May 2021. The diploma was structured in 7 modules and 40 Peruvian professionals participated, of which 11 (27.5%) were women, and 16 (40%) came from regions outside of Lima and Callao. We discussed the need to strengthen institutional and health professionals’ capacity to adequately manage large volumes of data, information, and knowledge through the application of emerging technologies to optimize data management processes to improve decision-making in health.