Causal Models and Big Data Learning Analytics

Recent critiques of both the uses of and discourse surrounding big data have raised important questions as to the extent to which big data and big data techniques should be embraced. However, while the context-dependence of data has been recognized, there remains a tendency among social theorists and other commentators to treat certain aspects of the big data phenomenon, including not only the data but also the methods and tools used to move from data as database to data that can be interpreted and assigned meaning, in a homogenizing way. In this paper, we seek to challenge this tendency, and to explore the ways in which explicit consideration of the plurality of big data might inform particular instances of its exploitation. We compare one currently popular big data-inspired innovation — learning analytics — with three other big data contexts — the physical sciences, business intelligence and public health. Through these comparisons, we highlight some dangers of learning analytics implemented without substantial theoretical, ethical and design effort. In so doing, we also highlight just how plural data, analytical approaches and intentions are, and suggest that each new big data context needs to be recognized in its own singularity.

Research background: Higher education institutions are generating multiple formats of data from diverse sources across the globe. The data ingestion layer is responsible for collecting data and transform for analysis. Learning analytics plays a vital role in providing decision-making support and selection of suitable timely intervention. The lack of tailored big-data ingestion architectures for academics led to several implementation challenges. Purpose of the article: The purpose of this article is to propose data ingestion architecture enabled for big data learning analytics. Methods: The study reviews existing literature to examine big-data ingestion tools and frameworks; and identify big-data ingestion challenges. An optimized framework for the real world learning analytics application was not yet in place at global higher educations. Consequently, the big-data ingestion pipeline is experiencing challenges of inefficient and complex data access, slow processing time, and security issues associated with transferring data to the system. The proposed data ingestion architecture is based on review of recent literature and adapts best international practices, guidelines, and techniques to meet the demand of current big-data ingestion issues. Findings & value added: This study identifies the current global challenges in implementing learning analytics projects. Review of recent big data ingestion techniques has been done based on defined metrics tuned for learning analytics purposes. The proposed data ingestion framework would increase the effectiveness of collecting, importing, processing and storing of learning data. Besides, the proposed architecture contributes to the construction of full-fledged big-data learning analytics ecosystem of higher educations.

PurposeAlthough big data analytics (BDA) have great benefits for higher education institutions (HEIs), due to lack of sufficient evidence on how BDA investment can pay off, it is tough for HEIs practitioners to realize value from such adoption. The purpose of this paper is to propose a big data academic and learning analytics enabled business value model to explain BDA potential benefits and business value which can be obtained by developing such analytics capabilities in HEIs.Design/methodology/approachThe study examined 47 case descriptions from 26 HEIs to investigate the causal association between the BDA current and potential benefits and business value creation path for big data academic and learning analytics success in HEIs.FindingsThe pressure of compliance with all legal and regulatory requirements and competition had pushed HEIs hard to adopt BDA tools. However, the study found out that application of risk and security and predictive analytics to higher education fields is still in its infancy. Using this theoretical model, the results provide new insights to higher education administrators on ways to create BDA capabilities for HEIs transformation and suggest an empirical foundation that can lead to more thorough analysis of BDA implementation.Originality/valueA distinctive theoretical contribution of this study is its conceptualization of understanding business value from BDA in the typical setting of higher education. The study provides HEIs with an all-inclusive understanding of BDA and gives insights on how it helps to transform HEIs. The new perspectives associated with the big data academic and learning analytics enabled business value model will contribute to future research in this area.

The development of mobile Internet, Internet of Things, and cloud computing has contributed to the unprecedented growth of information data. Big data plays a very important role in education. Currently, the literature review and in-depth research on big educational data are not very extensive, mainly involved in two fields: education mining and learning analysis. For a perfect research about education big data, this paper comprehensively reviewed three major aspects (Predictive Analytics, Learning Analytics, and Recommendation Systems) of educational data analytics for an intensive investigation and analysis: (1) Predictive Analytics: It predicts students’ learning performance by tracking students’ learning information and then analyzes students’ learning competence to build an academic early warning system; teachers can be allowed to intervene in students in time and adopts different teaching ways for different students. Therefore, both students’ learning and ability can be individualized and improved; (2) Learning Analytics: This part can identify the learners’ behavior patterns and obtain more implicit learner characteristics by studying the hidden meaning behind learning behaviors and strategies; (3) Recommendation Systems: It can match the needs of learners and recommend appropriate learning resources through different methods. All the above proved that the application of big data technology in education provides powerful data support for the development of education.

New statistical methods allow discovery of causal models from observational data in some circumstances. These models permit both probabilistic and causal inference for models of reasonable size. Many domains can benefit from such methods. Educational research does not easily lend itself to experimental investigation. Research in laboratories is artificial while research in authentic environments is complex and difficult to control. The variables are typically hidden and change over the long term, making them challenging and expensive to investigate experimentally. We present an analysis of causal discovery algorithms and their applicability to educational research and learning technology, an engineered causal model of self-regulated learning (SRL) theory based on the literature, and an evaluation of the potential for discovering such a model from observational data using the new statistical methods.

With the arrival of 'big data; in education, the potential was recognised for learning analytics to track students' learning, to reveal patterns in their learning, or to identify at-risk students, in addition to guiding reform and supporting educators in improving teaching and learning processes [1]. Learning Analytics dashboards have been used at all levels, including institutional, regional and national level [2]. In classroom use, while learning visualisations are often based on counts of activity data or interaction patterns, there is increasing recognition that learning analytics relate to learning, and should therefore provide pedagogically useful information [3]. While increasing numbers of technology-enhanced learning applications are embracing the potential of learning analytics at the classroom level, often these are aimed at teachers. However, learners can also benefit from learning analytics data (e.g. [4][5]).

The exploding number of computational models produced by Systems Biologists over the last years is an invitation to structure and exploit this new wealth of information. Researchers would like to trace models relevant to specific scientific questions, to explore their biological content, to align and combine them, and to match them with experimental data. To automate these processes, it is essential to consider semantic annotations, which describe their biological meaning. As a prerequisite for a wide range of computational methods, we propose general and flexible similarity measures for Systems Biology models computed from semantic annotations. By using these measures and a large extensible ontology, we implement a platform that can retrieve, cluster, and align Systems Biology models and experimental data sets. At present, its major application is the search for relevant models in the BioModels Database, starting from initial models, data sets, or lists of biological concepts. Beyond similarity searches, the representation of models by semantic feature vectors may pave the way for visualisation, exploration, and statistical analysis of large collections of models and corresponding data.

Tremendous proliferation in data generation in the past few years has paved the way for new research and the development of new and improved techniques and algorithms in different fields of science and education. Initially terms like educational data mining emerged as a branch of data mining borrowing techniques from its ancestor. The challenges brought about by this large and heterogeneous data are diverse and needs a greater serious technical treatment. New and emerging fields like learning analytics have been introduced to manage the complexities of this data deluge. Learning analytics deals with data in the context of learner and the learning environment to improve the overall learning experience. The ultimate aim of the field is to make use of the data about learners and their environments to gain insights into the learning process using some of the well-known techniques and algorithms from the fields of data mining and machine learning. The process involves collecting, analysis of data and reporting the results to understand and optimize the learning experience. The fields of data mining and academic analytics closely related to learning analytics. Systematic Literature Review (SLR) is a robust, organized and rigorous literature review and reporting process aimed at identifying, collecting and synthesizing the relevant literature on a research question according to specified criteria. The process is more unbiased and balanced by systematic sequence of steps. This paper presents a systematic literature review by first developing the systematic literature review protocol and then discussing the main findings of the literature review by especially focusing on the applications and uses of machine learning and data mining techniques in the domain of learning analytics.
 
 Index Terms—Systematic Literature Review (SLR), Learning Analytics (LA), Big Data, Educational Data Mining (EDM), Machine Learning (ML).

The paper describes transformation rules that map between IDEFO and IDEF3 integration definitions and Petri net modelling constructs. Use of these rules has enabled the development of an enterprise modelling methodology. This methodology ensures that different yet consistent views of business processes and their underpinning resources can be developed by members of a systems engineering team. In particular, the methodology supports the organized capture of functional models, expressed in IDEFO notation, and links them to the subsequent development of behaviour models expressed in IDEF3 notation. Decomposition rules have also been developed that determine relationships between system elements so that they can be modelled at various levels of granularity. Resultant behaviour models can be translated into an equivalent Petri net form. One target application of the modelling rules and methods has been to model interactions between business partners that share common business processes.

Recent years have seen a growing recognition of the value of positioning teaching as a design science. In this context, we see a great potential in the possible synergy between educational design, learning analytics and teacher inquiry. This synergy is demonstrated in the form of C omputer S upported C urriculum A nalysis, D esign and E valuation for S cience E ducation in A frica ( CASCADE ‐ SEA ), a computer‐based support tool that assists teachers in creating exemplary lesson materials for secondary level science and mathematics in southern A frica. The purpose of the present study is to articulate how teachers might be supported in the synergistic processes that are integral to educational design. It does so by revisiting existing data collected during the design, development and evaluation of CASCADE ‐ SEA , and sharing the results of a retrospective analysis in light of recent developments in learning analytics, learning design and teacher inquiry. The retrospective analysis revealed insights related to each of the aforementioned themes. For learning analytics, those were social structures of learning and teaching practices; pedagogy‐driven collection and analysis of data; and the use of data to inform teaching and learning. For learning design, the findings pertained to the language, practices and tools used to achieve particular educational aims. For teacher inquiry, the findings relate to the empirical process of question formulation, methods selection, collection and analysis of data, and sharing/reflecting on the findings as well as overall process. Based on the retrospective analysis, design considerations are offered for a modern application to replace the practical and effective desktop software that was developed during the 1990s.

Data-informed teacher inquiry into their own practices should help teachers pick most effective teaching methods for each particular group of students and teaching situation. However, there is not much research into the thought processes that teachers employ when making these decisions. This study is an attempt at finding out how data about teaching and learning from physical classrooms, collected with digital tools, could help teachers in adjusting their teaching methods and strategies to improve practice. The study explores how 12 Estonian teachers perceive their own technology use and teacher inquiry (TI) practices, and how much this corresponds to reality. Also, how data-analysis skills satisfy their actual needs and what problems they encounter within a technology-enhanced TI process. Survey data and TI diaries are analyzed using descriptive statistics and qualitative content analysis, and the teachers’ perceived knowledge of technology and teacher inquiry is compared to their actual application of these in 18 interventions. In addition, thematic analysis is applied to TI diary entries about teachers’ sense-making and interpretation of the collected data to provide insights into teachers’ decision-making processes, which would enable improvements to teacher inquiry models and further inform the development of different teaching and learning analytics (TLA) tools. Finally, teachers’ concerns connected with different stages in TI facilitation are calculated to identify how these relate to teachers’ application of TI.

The unique position of learning analytics at the intersection of education and computer science while reaching out to several other disciplines such as statistics, psychometrics, econometrics, mathematics, and linguistics has accelerated the growth and expansion of the field. Therefore, it is a crucial endeavor for learning analytics researchers to stay abreast of the latest methodological and computational advances to drive their research forward. The diversity and complexity of the existing methods can make this task overwhelming both for newcomers to the learning analytics field and for experienced researchers. With the motivation to accompany researchers in this challenging journey, the book “Learning Analytics Methods and Tutorials—A Practical Guide Using R” aims to provide a methodological guide for researchers to study, consult, and take the first steps toward innovation in the learning analytics field. Thanks to the unique wealth of authors’ backgrounds and expertise, which include authors of R packages and experts in methods and applications, the book offers a comprehensive array of methods that are described thoroughly with a primer on their usage in prior research in education. These methods include sequence analysis, Markov models, factor analysis, process mining, network analysis, predictive modeling, and cluster analysis among others. A step-by-step tutorial using the R programming language with real-life datasets and case studies is presented for each method. In addition, the initial chapters are devoted to getting novice researchers up to speed with the R programming learners and the basics of data analysis. The present chapter serves as an introduction to the book describing its main aim and intended audience. It describes the structure of the book and the methods covered by each chapter. It also points the readers to the companion code and data repositories to facilitate following the tutorials present in the book chapter.

Any implementation of personalized learning and teaching through big data will have implications for research ethics. This chapter considers the research ethics implications in certain key regulatory documents and the scholarship on research ethics and learning analytics. Finally, the chapter provides guidelines for use by researchers and research ethics review committees—within the field of education in the South African context—specifically focusing on personalized learning and teaching through big data and learning analytics.

: Despite significant proliferation of Internet services in recent years, technology for computer-supported cooperative work and groupware have not progressed at the same rate. A wider distribution of the work force motivates the need for networked multimedia and groupware at Internet scope and for larger groups of end-users. In particular, synchronous telecollaboration enables people in different geographic locations to bridge time and space by sharing and jointly manipulating multimedia information in realtime and at various levels of granularity. This aspect stands in contrast to legacy client-server applications such as Internet radio broadcast or video-on-demand, and to asynchronous, document-centric collaboration tools like email, instant messaging, or chat rooms. In this paper, we provide a framework for network-supported synchronous multimedia groupwork at Internet scope and for large user groups. Contributions entail an novel classification for such systems concerning scale and scope of interaction, a formal framework for Internet sessions and mediation of access to concurrently shared resources, a taxonomy of crucial elements in cooperative applications, and a discussion of a generic network coordination protocol to sustain live interaction among concurrently active user groups. The core ideas put forward in this paper are useful for the characterization and rapid prototyping of a new generation of collaborative applications.

In this article, we investigate the role of gender in collaboration patterns by analyzing gender-based homophily-the tendency for researchers to co-author with individuals of the same gender. We develop and apply novel methodology to the corpus of JSTOR articles, a broad scholarly landscape, which we analyze at various levels of granularity. Most notably, for a precise analysis of gender homophily, we develop methodology which explicitly accounts for the fact that the data comprises heterogeneous intellectual communities and that not all authorships are exchangeable. In particular, we distinguish three phenomena which may affect the distribution of observed gender homophily in collaborations: a structural component that is due to demographics and non-gendered authorship norms of a scholarly community, a compositional component which is driven by varying gender representation across sub-disciplines and time, and a behavioral component which we define as the remainder of observed gender homophily after its structural and compositional components have been taken into account. Using minimal modeling assumptions, the methodology we develop allows us to test for behavioral homophily. We find that statistically significant behavioral homophily can be detected across the JSTOR corpus and show that this finding is robust to missing gender indicators in our data. In a secondary analysis, we show that the proportion of women representation in a field is positively associated with the probability of finding statistically significant behavioral homophily.