Data Science Methods Research Articles

Hidden Markov Model Segmentation to Demarcate Trajectories of Residual Apnoea-Hypopnoea Index in CPAP-Treated Sleep Apnoea Patients: The New Concept of a Telemonitogram

Background: Continuous positive airway pressure (CPAP), the reference treatment for obstructive sleep apnoea (OSA), is used by millions of individuals worldwide with remote telemonitoring providing daily information on CPAP usage and efficacy, a currently underused resource. Here, we aimed to implement state-of-the-art data science methods to describe heterogeneity and diversity of time-series of residual apnoea-hypopnoea indexes (rAHI) from CPAP telemonitoring. Methods: We analysed a CPAP telemonitoring database to model and cluster rAHI trajectories. Our primary objective was to use Hidden Markov models (HMMs) as a probabilistic model-based approach to extract features from rAHI time-series. Secondary goals were to identify clusters of rAHI trajectories and their relation to CPAP treatment outcomes, adherence and leaks. Findings: From telemonitoring records of 2,860 CPAP-treated patients (age: 66·31 ± 12·92 years, 69·9% male), HMM modelling revealed three states differing in variability within a given state and probability of shifting from one state to another. Six clusters of rAHI trajectories were identified ranging from well controlled CPAP-treated patients (Cluster 0: 669 (23%); mean rAHI 0·58 ± 0·59 events/hour) to the most unstable (Cluster 5: 470 (16%); mean rAHI 9·62 ± 5·62 events/hour). CPAP adherence was 30 minutes higher in cluster 0 compared to clusters 4 and 5 (p-value <0·01). Leaks were significantly more frequent in cluster 5. Interpretation: This new approach based on HMM might constitute the backbone for deployment of digital health solutions improving the interpretation of telemonitoring data from CPAP-treated patients. Funding Information: JLP, SB, and RT are supported by the French National Research Agency in the framework of the Investissements d’avenir” program (ANR-15-IDEX-02) and the “e-health and integrated care and trajectories medicine and MIAI artificial intelligence” Chairs of excellence from the Grenoble Alpes University Foundation (ANR-19-P3IA-0003). AM is supported by Probayes and MIAI in the framework of a “Convention Industrielle de Formation par la Recherche” (CIFRE) PhD. The PhD is also supported by the French National Research Agency (grant 2020/0007). Declaration of Interests: JLP reports grants from Air Liquide Foundation, grants and personal fees from Agiradom, grants and personal fees from AstraZeneca, grants from Fisher and Paykel, grants from Mutualia, grants and personal fees from Philips, grants and personal fees from Resmed, grants from Vitalaire, grants from Boehringer Ingelheim, grants from Jazz Pharmaceuticals, grants from Night Balance, grants from Sefam, outside the submitted work; JCB is employee of AGiR-a-dom, a French home-care provider; RT reports other from AGiR-a-dom, grants from Resmed, outside the submitted work; AM, RLH and MCS report other from AGiR-a-dom, and SB have no conflict of interest directly related to this work. Ethics Approval Statement: We analysed a CPAP telemonitoring database [registered and ethically approved by the French C.C.T.I.R.S: N°15.925bis and ethics regulations MR003 N° 1996650v0] to model and cluster patients’ rAHI trajectories.

FRAMEWORK FOR SYSTEMATIZATION OF DATA SCIENCE METHODS

The rapid development of data science has led to the accumulation of many models, methods, and techniques that had been successfully applied. As the analysis of publications has shown, the systematization of data science methods and techniques is an urgenttask. However, in most cases, the results are relevant to applications in a particular problem domain. The paper develops the framework for the systematization of data science methods, neither domain-oriented nor task-oriented. The metamodel-method-technique hierarchy organizes the relationships between existing methods and techniques and reduces the complexity of their under-standing. The first level of the hierarchy consists of metamodels of data preprocessing, data modeling, and data visualization. The second level comprises methods corresponded to metamodels. The third level collects the main techniques grouped according to methods.The authors describe the guiding principles of the framework use. It provides a possibility to define the typical process of problem-solving with data science methods. A case study is used to verify the framework’s appropriateness. Four cases of applying data science methods to solve practical problems described in publications are examined. It is shown that the described solutions are entirely agreed with the proposed framework. The recommended directions for applying the framework are defined. The constraint of the framework applying is structured or semi-structured data that should be analyzed. Finally, the ways of further research are given.

Data science for engineering design: State of the art and future directions

Engineering design (ED) is the process of solving technical problems within requirements and constraints to create new artifacts. Data science (DS) is the inter-disciplinary field that uses computational systems to extract knowledge from structured and unstructured data. The synergies between these two fields have a long story and throughout the past decades, ED has increasingly benefited from an integration with DS.We present a literature review at the intersection between ED and DS, identifying the tools, algorithms and data sources that show the most potential in contributing to ED, and identifying a set of challenges that future data scientists and designers should tackle, to maximize the potential of DS in supporting effective and efficient designs. A rigorous scoping review approach has been supported by Natural Language Processing techniques, in order to offer a review of research across two fuzzy-confining disciplines.The paper identifies challenges related to the two fields of research and to their interfaces. The main gaps in the literature revolve around the adaptation of computational techniques to be applied in the peculiar context of design, the identification of data sources to boost design research and a proper featurization of this data. The challenges have been classified considering their impacts on ED phases and applicability of DS methods, giving a map for future research across the fields. The scoping review shows that to fully take advantage of DS tools there must be an increase in the collaboration between design practitioners and researchers in order to open new data driven opportunities.

Urban Air Pollution Mapping Using Fleet Vehicles as Mobile Monitors and Machine Learning.

Spatially explicit urban air quality information is important for developing effective air quality control measures. Traditionally, urban air quality is measured by networks of stationary monitors that are not universally available and sparsely sited. Mobile air quality monitoring using equipped vehicles is a promising alternative but has focused on vehicle-level experiments and lacks fleet-level demonstration. Here, we equipped 260 electric vehicles in a ride-hailing fleet in Beijing, China with low-cost sensors to collect real-time, spatial-resolved data on fine particulate matter (PM2.5) concentrations. Using this data, we developed a decision tree model to infer the distribution of PM2.5 concentrations in Beijing at 1 km by 1 km and 1 h resolution. Our results are able to show both short- and long-term variations of urban PM2.5 concentrations and identify local air pollution hotspots. Compared with a benchmark model that only uses data from stationary monitoring sits, our model has shown significant improvement with the coefficient of determination increased from 0.56 to 0.80 and root mean square error decreased from 12.6 to 8.1 μg/m3. To the best of our knowledge, this study collects the largest mobile sensor data for urban air quality monitoring, which are augmented by state-of-the-art machine learning techniques to derive high-quality urban air pollution mapping. Our results demonstrate the potential and necessity of using fleet vehicles as routine mobile sensors combined with advanced data science methods to provide high-resolution urban air quality monitoring.

Assessing the educational performance of different Brazilian school cycles using data science methods.

Educational indicators are metrics that assist in assessing the quality of the educational system. They are often associated with economic and social factors suggested to contribute to good school performance, however there is no consensus on the impact of these factors. The main objective of this work was to evaluate the factors related to school performance. Using a data set composed by Brazilian schools’ performance (IDEB), socioeconomic and school structure variables, we generated different models. The non-linear model predicted the best performance, measured by the error and determination coefficient metrics. The heterogeneity of the importance of the variable between school cycles and regions of the country was detected, this effect may contribute to the development of public educational policies.

How Hydrogen and Oxygen Vapor Affect the Tribochemistry of Silicon- and Oxygen-Containing Hydrogenated Amorphous Carbon under Low-Friction Conditions: A Study Combining X-ray Absorption Spectromicroscopy and Data Science Methods.

The incorporation of silicon and oxygen into hydrogenated amorphous carbon (a-C:H) is an effective approach to decrease the dependence of the tribological properties of a-C:H on the environment. Here, we evaluate the effect of hydrogen and oxygen partial pressures in vacuum on the tribological response of steel pins sliding against films consisting of silicon- and oxygen-containing a-C:H (a-C:H:Si:O). Experiments are conducted in the low-friction/low-wear regime, where sufficient gas pressure prevents steel from adhering to the a-C:H:Si:O, with the velocity accommodation mode being interfacial sliding between the tribotrack formed in the a-C:H:Si:O film and the carbonaceous tribofilm that is formed on the countersurface. The experiments indicated a decrease (increase) in friction and wear with the hydrogen (oxygen) pressure (hydrogen pressures between 50 and 2000 mbar; oxygen pressures between 10 and 1000 mbar). Characterization by X-ray photoelectron and absorption spectroscopies indicated the occurrence of tribologically induced rehybridization of carbon-carbon bonds from sp3 to sp2. This mechanically induced structural transformation coincided with the dissociative surface reaction between hydrogen (oxygen) gas molecules and sp2 carbon-carbon bonds that are highly strained, which results in the formation of carbon-hydrogen groups (carbonyl or ether groups together with silicon atoms having higher oxidation states). On the basis of variations of the fraction of these surface functional groups with gas pressure, a phenomenological model is proposed for the gas pressure dependence of friction for steel when sliding on a-C:H:Si:O films: while the decrease in friction with hydrogen pressure is induced by an increase in the percentage of carbon-hydrogen groups, the increase in friction with oxygen pressure is caused by a progressive increase in the relative fraction of silicon atoms having higher oxidation states and an increase in surface oxygen concentration.

Application of Predictive Methods to Financial Data Sets

Abstract Financial data sets are growing too fast and need to be analyzed. Data science has many different techniques to store and summarize, mining, running simulations and finally analyzing them. Among data science methods, predictive methods play a critical role in analyzing financial data sets. In the current paper, applications of 22 methods classified in four categories namely data mining and machine learning, numerical analysis, operation research techniques and meta-heuristic techniques, in financial data sets are studied. To this end, first, literature reviews on these methods are given. For each method, a data analysis case (as an illustrative example) is presented and the problem is analyzed with the mentioned method. An actual case is given to apply those methods to solve the problem and to choose a better one. Finally, a conclusion section is proposed.

A Multiple Subspaces-Based Model: Interpreting Urban Functional Regions with Big Geospatial Data

Analyzing the urban spatial structure of a city is a core topic within urban geographical information science that has the ability to assist urban planning, site selection, location recommendation, etc. Among previous studies, comprehending the functionality of places is a central topic and corresponds to understanding how people use places. With the help of big geospatial data which contain affluent information about human mobility and activity, we propose a novel multiple subspaces-based model to interpret the urban functional regions. This model is based on the assumption that the temporal activity patterns of places lie in a high-dimensional space and can be represented by a union of low-dimensional subspaces. These subspaces are obtained through finding sparse representations using the data science method known as sparse subspace clustering (SSC). The paper details how to use this method in the context of detecting functional regions. With these subspaces, we can detect the functionality of urban regions in a designated study area and further explore the characteristics of functional regions. We conducted experiments using real data in Shanghai. The experimental results and outperformance of our proposed model against the single subspace-based method prove the efficacy and feasibility of our model.

Analysis of Tumor Disease Patterns Based on Medical Big Data

Using the medical big data mining related technology, the model of tumor disease was analyzed and studied. Using data science methods as a guiding method and idea, analyzing and constructing a medical service model based on big data for oncology diseases, exploring its development strategy; using business process analysis method to analyze the business process and mapping of cancer disease medical services; using serviceoriented architecture analysis and Design methodology to build a highly flexible, configurable, and easily scalable precision medical big data platform. By analyzing the characteristics of medical big data and the shortcomings of the traditional Apriori algorithm, the Hadoop platform is used to improve and optimize the Apriori algorithm. The results show that the improved Apriori algorithm has great improvement in efficiency and performance, and can be adapted to mining medical big data. Through data mining experiments, it is concluded that there is a correlation between tumors and smoking, chronic infection, occupational pathogenic factors, etc. It has certain guiding significance for the prevention and treatment of tumors, thus also demonstrating the improved Apriori algorithm for lung tumors. Clinical research has practical significance.

On the Use of Topological Features of Metabolic Networks for the Classification of Cancer Samples.

Background The increasing availability of omics data collected from patients affected by severe pathologies, such as cancer, is fostering the development of data science methods for their analysis.Introduction The combination of data integration and machine learning approaches can provide new powerful instruments to tackle the complexity of cancer development and deliver effective diagnostic and prognostic strategies.Methods We explore the possibility of exploiting the topological properties of sample-specific metabolic networks as features in a supervised classification task. Such networks are obtained by projecting transcriptomic data from RNA-seq experiments on genome-wide metabolic models to define weighted networks modeling the overall metabolic activity of a given sample.Results We show the classification results on a labeled breast cancer dataset from the TCGA database, including 210 samples (cancer vs. normal). In particular, we investigate how the performance is affected by a threshold-based pruning of the networks by comparing Artificial Neural Networks, Support Vector Machines and Random Forests. Interestingly, the best classification performance is achieved within a small threshold range for all methods, suggesting that it might represent an effective choice to recover useful information while filtering out noise from data. Overall, the best accuracy is achieved with SVMs, which exhibit performances similar to those obtained when gene expression profiles are used as features.Conclusion These findings demonstrate that the topological properties of sample-specific metabolic networks are effective in classifying cancer and normal samples, suggesting that useful information can be extracted from a relatively limited number of features.

Artificial Intelligence and Telehealth may Provide Early Warning of Epidemics.

The COVID-19 pandemic produced a very sudden and serious impact on public health around the world, greatly adding to the burden of overloaded professionals and national medical systems. Recent medical research has demonstrated the value of using online systems to predict emerging spatial distributions of transmittable diseases. Concerned internet users often resort to online sources in an effort to explain their medical symptoms. This raises the prospect that incidence of COVID-19 may be tracked online by search queries and social media posts analyzed by advanced methods in data science, such as Artificial Intelligence. Online queries can provide early warning of an impending epidemic, which is valuable information needed to support planning timely interventions. Identification of the location of clusters geographically helps to support containment measures by providing information for decision-making and modeling.

Teaching Data Science in Political Science: Integrating Methods with Substantive Curriculum

ABSTRACTThe importance of data science in society today is undeniable, and now is the time to prepare data science talent (National Academies of Sciences, Engineering, and Medicine 2018). Data science demands collaboration, but collaboration within political science departments has been weak in teaching data science. Bridging substantive and methods courses can critically aid in teaching data science because it facilitates this collaboration. Our innovation is to integrate data science into both substantive and methods courses through a dedicated data science course and modules on data science topics taught in substantive courses. This approach allows not only for more opportunities for teaching and practice of data science methods but also helps students to understand how social, economic, and political biases and incentives can affect their data.

Systematic Mapping of Global Research on Climate and Health Using Machine Learning

Background: The global literature on the links between climate change and human health is large, increasing exponentially, and no longer feasible to collate and synthesize using traditional systematic review approaches. Methods: We use machine learning and other data science methods to systematically identify and map the scientific literature on climate change and health published since 2013. We classify >16,000 relevant publications by field of climate research, climate drivers, health impact, date and geography. Findings: We show that climate-health literature is dominated by impact studies, with mitigation and adaptation responses and their co-benefits and co-risks remaining niche topics. Air quality and heat stress are the most frequently studied exposures, with all-cause mortality and infectious disease incidence being the most frequently studied health outcomes. Seasonality, extreme weather events, heat, and weather variability are the most frequently studied climate-related hazards. Geographically, the evidence base is dominated by studies from high-income countries and China, with very limited evidence from the majority of countries that suffer most from the health consequences of climate change. Interpretation: We demonstrate the importance and feasibility of using automated machine-learning pipelines for mapping the science on climate change and human health and generating valuable inputs into global climate and health assessments. We find major gaps in evidence on climate-health research for mental health, undernutrition, and maternal-child health, and very limited evidence overall from low income regions. The scant evidence on climate change response options is concerning and will significantly hamper the design of evidence-based pathways to reduce the effects on health of climate change. We argue that in the post-2015 Paris Agreement era of climate solutions much more attention should be given to climate adaptation and mitigation options and their effects on human health. Funding Statement: None to declare. Declaration of Interests: Dr. Berrang-Ford reports grants from Dfid-FCDO UK (Gov't), from null, during the conduct of the study; all other authors have nothing to declare.

Application of data science methods to identify school and home risk factors for asthma and allergy-related symptoms among children in New York

ObjectivesFew studies have comprehensively assessed multiple environmental exposures affecting children's health. This study applied machine-learning methods to evaluate how indoor environmental conditions at home and school contribute to asthma and allergy-related symptoms. MethodsWe randomly selected 10 public schools representing different socioeconomic statuses in New York State (2017–2019) and distributed questionnaires to students to collect health status and home-and school-environmental exposures. Indoor air quality was measured at school, and ambient particle exposures (PM2.5 and components) were measured using real-time personal monitors for 48 h. We used random forest model to identify the most important risk factors for asthma and allergy-related symptoms, and decision tree for visualizing the inter-relationships among the multiple risk factors with the health outcomes. ResultsThe top contributing factors identified for asthma were family rhinitis history (relative importance: 10.40%), plant pollen trigger (5.48%); bedroom carpet (3.58%); environmental tobacco smoke (ETS) trigger symptom (2.98%); and ETS exposure (2.56%). For allergy-related symptoms, plant pollen trigger (10.88%), higher paternal education (7.33%), bedroom carpet (5.28%), family rhinitis history (4.78%), and higher maternal education (4.25%) were the strongest contributing factors. Conversely, primary heating with hot water radiator was negatively (−6.86%) associated with asthma symptoms. Younger children (<9 years old) with family history of rhinitis and carpeting in the bedroom were the prominent combined risk factors for asthma. Children jointly exposed to pollen, solvents, and carpeting in their home tended to have greater risks of allergy-related symptoms, even without family history of rhinitis. ConclusionFamily rhinitis history, bedroom carpet, and pollen triggers were the most important risk factors for both asthma and allergy-related symptoms. Our new findings included that hot-water radiator was related to reduced asthma symptoms, and the combination of young age, rhinitis history, and bedroom carpeting was related to increased asthma symptoms. Further studies are needed to confirm our findings.

Quantum computing at the frontiers of biological sciences.

The search for meaningful structure in biological data has relied on cutting-edge advances in computational technology and data science methods. However, challenges arise as we push the limits of scale and complexity in biological problems. Innovation in massively parallel, classical computing hardware and algorithms continues to address many of these challenges, but there is a need to simultaneously consider new paradigms to circumvent current barriers to processing speed. Accordingly, we articulate a view towards quantum computation and quantum information science, where algorithms have demonstrated potential polynomial and exponential computational speedups in certain applications, such as machine learning. The maturation of the field of quantum computing, in hardware and algorithm development, also coincides with the growth of several collaborative efforts to address questions across length and time scales, and scientific disciplines. We use this coincidence to explore the potential for quantum computing to aid in one such endeavor: the merging of insights from genetics, genomics, neuroimaging and behavioral phenotyping. By examining joint opportunities for computational innovation across fields, we highlight the need for a common language between biological data analysis and quantum computing. Ultimately, we consider current and future prospects for the employment of quantum computing algorithms in the biological sciences.

Fixed Point Strategies in Data Science

The goal of this paper is to promote the use of fixed point strategies in data science by showing that they provide a simplifying and unifying framework to model, analyze, and solve a great variety of problems. They are seen to constitute a natural environment to explain the behavior of advanced convex optimization methods as well as of recent nonlinear methods in data science which are formulated in terms of paradigms that go beyond minimization concepts and involve constructs such as Nash equilibria or monotone inclusions. We review the pertinent tools of fixed point theory and describe the main state-of-the-art algorithms for provably convergent fixed point construction. We also incorporate additional ingredients such as stochasticity, block-implementations, and non-Euclidean metrics, which provide further enhancements. Applications to signal and image processing, machine learning, statistics, neural networks, and inverse problems are discussed.

World Heritage sites on Wikipedia: Cultural heritage activism in a context of constrained agency

UNESCO World Heritage sites are places of outstanding significance and often key sources of information that influence how people interact with the past today. The process of inscription on the UNESCO list is complicated and intersects with political and commercial controversies. But how well are these controversies known to the public? Wikipedia pages on these sites offer a unique dataset for insights into public understanding of heritage controversies. The unique technicity of Wikipedia, with its bot ecosystem and editing mechanics, shapes how knowledge about cultural heritage is constructed and how controversies are negotiated and communicated. In this article, we investigate the patterns of production, consumption, and spatial and temporal distributions of Wikipedia pages for World Heritage cultural sites. We find that Wikipedia provides a distinctive context for investigating how people experience and relate to the past in the present. The agency of participants is highly constrained, but distinctive, behind-the-scenes expressions of cultural heritage activism are evident. Concerns about state-like actors, violence and destruction, deal-making, etc. in the World Heritage inscription process are present, but rare on Wikipedia’s World Heritage pages. Instead, hyper-local and process issues dominate controversies on Wikipedia. We describe how this kind of research, drawing on Big Data and data science methods, contributes to digital heritage studies and also reveals its limitations.

Food Demand Prediction Using the Nonlinear Autoregressive Exogenous Neural Network

Food demand prediction is a significant issue for both business process improvement and sustainable development. Data science methods, including artificial intelligence methods, are often used for this purpose. The aim of this research is to develop models for food demand prediction based on a nonlinear autoregressive exogenous neural network. The research focuses on processed foods, such as bread or butter. The architectures of the developed models differed in the number of hidden layers and the number of neurons in the hidden layers, as well as with different sizes of the delay line, were tested for a given product. The results of the research show that depending on the type of product, the prediction performance slightly differed. The results of the R2 measure ranged from 96,2399 to 99,6477, depending on particular products. The proposed models can be used in a company’s intelligent management system for the rational control of inventories and food production. This can also lead to a reduction in food waste.

Concept for Enabling Customer-oriented Data Analytics via Integration of Production Process Improvement Methods and Data Science Methods

Production companies are facing the major challenge of digitization. New technical developments enable new optimization potentials in production that can be leveraged in a highly competitive market. Data analysis is a good example of this, allowing large amounts of data to be used to support production in optimization projects. For data analysis the CRISP-DM approach has become generally accepted in science and practice. This process model offers a good support for data science projects with useful data analysis methods and tools. In practice, however, it can often be observed that aspects such as customer-oriented project definition and the integration of process knowledge in data science projects are difficult to achieve. This paper will present a solution for that challenge. The combination of process optimization methods from the Lean Management and Six Sigma toolbox as well as project management methods for each phase of the CRISP-DM approach support data science projects to customize the project definition and integrate process knowledge.

Research on River Governance Index System in the Big Data Era

Realizing the ecological health and sustainable development of rivers has become a worldwide research hotspot. Sustainable development is a large system with multi-system, multi-level and multi-index, In the era of big data, data science methods have made great changes in water resources governance methods. We can obtain more detailed data through big data methods to comprehensively and scientifically measure river governance. Based on the analytic hierarchy process, this paper invites experts in related fields to construct and score the river governance index system, and establish a river governance index system in the era of big data. We hope that the river governance index system will achieve advanced river governance in the era of big data.