Search

Various computing subdisciplines, such as computer science and software engineering, each have their own curricular guidelines. They can be very difficult to understand and compare for people such as prospective students, industry personnel, and even faculty members. This is compounded by a lack of information surrounding undergraduate computing curricular topics via visual methods. This paper describes two experimental activities where the objective is to explore the possibility of obtaining quantitative data sets necessary for visualization, one based on competencies and the other based on knowledge areas. Both activities were based on surveys. The results from the first activity showed that a consensus interpretation could be obtained for the knowledge, skills, and dispositions implied by the competency descriptions, although not as strongly for dispositions. The second activity resulted in a table of knowledge areas with minimum and maximum weights for six computing subdisciplines. Finally, this paper also shows two examples of how users can explore the various curricular guidelines through visualization.

Automatic Indexing of documents using paragraph vectors is a popular unsupervised method for learning distributed representations of texts. This method learns embedding of words with document vectors for document classification. In addition, this method can be leveraged for sentiment analysis. However, while the results presented in the original Doc2Vec study were promising, the overall proof of concept was rather narrow. In this study, we extend the Doc2Vec method, and enhance it to classify locality sensitive documents, i.e. domain-based documents which are largely similar with marginal differences. In particular, we use the enhanced Doc2Vec technique to classify similar documents describing the Internet of Things (IoT) patterns. We observed that our enhanced locality sensitive Doc2Vec technique performs significantly well to improve embedding quality. The model performance is in par with state-of-the-art results and can be qualified as a benchmark for similar vector space models.

The paper presents a systematic literature review of the literature on the competencies that are essential to develop a globally competitive workforce in the field of data science. The systematic review covers a wide range of literature but focuses primarily, but not exclusively, on the computing, information systems, management, and organisation science literature. The paper uses a broad research search strategy covering four separate electronic databases. The search strategy led the researchers to scan 139 titles, abstracts and keywords. Sixty potentially relevant articles were identified, of which 42 met the quality criteria and contributed to the analysis. A critical appraisal checklist assessed the validity of each empirical study. The researchers grouped the findings under six broad competency themes: organisational, technical, analytical, ethical and regulatory, cognitive and social. Thematic analysis was used to develop a unified model of data science competency based on the evidence of the findings. This model will be applied to case studies and survey research in future studies. A unified data science competency model, supported by empirical evidence, is crucial in closing the skills gap, thereby improving the quality and competitiveness of the South Africa's data science workforce. Researchers are encouraged to contribute to the further conceptual development of data science competency.

The Internet of Things (IoT) is a paradigm with multitudes of design patterns. However, in order to use these patterns quickly and effectively, one must be able to make a differentiation between the existing patterns. At the moment, there is no known catalogue for the IoT patterns in which each pattern is described at a fine-grained level, i.e. in terms of its attributes. The need to discuss these patterns in terms of their attributes is important as it enables ease of understanding and allows us to group related patterns together for speedy retrieval. In this paper, we present an attributes extraction system which generates a list of attributes for a given IoT pattern. The attributes extraction system is based on identification and extraction of important sentences which describe the core properties of the given IoT pattern. The system uses multiple linguistics features to identify the most important sentences in a document with regard to describing the core essence of a given pattern. The system calculates an independent score for each sentence per feature. Through aggregation, the independent scores for each feature can then be combined to give a weighted mean score for each sentence. The evaluation results show that the attributes selected by the system are consistent with human ranking in the bulk of the examined documents.

Most popular industry verticals such as smart health, smart mining, smart agriculture, etc. fail to map clearly the solution patterns used in the Internet of Things (IoT). This causes a disconnection in our understanding of the implementation approaches and solutions used in the IoT paradigm. Nonetheless, there is another way of subdividing this space – decoding the building blocks of the IoT architectures, with a focus on common components and patterns for interoperability. The IoT presents us with several design challenges at all architectural levels. This ranges from the overall architecture at the highest level to device connectivity at the lowest level. In this complex design, it is easy to get lost in the forest of the latest IoT offerings without understanding the underlying solution patterns. In this study, we have mapped several IoT architectural patterns, spanning several industry verticals, by analysing the literature that report on real-world IoT implementations to identify common, recurring solution patterns. In the spirit of the Gang-of-Four 1 and Christopher Alexander’s pattern language, as well as some recent advances in the IoT space, we have identified some systematic methods which are used to organise the IoT patterns.

The study of algorithms which solve the subgraph isomorphism problem is very important because it has many applications where data is modelled as graphs. Despite the subgraph isomorphism problem being NP-hard, research has been dedicated to proposing new algorithms which are designed to improve the shortcomings of the algorithms that have been previously proposed. However, some of the newly invented solutions do not improve on their predecessors and some are redundant because their algorithmic properties are not different from some existing algorithms. The problem arises because minimal effort has been made towards organising subgraph isomorphism algorithms in terms of their relationships and differences such that new solutions utilise knowledge extracted from the organisation of knowledge about these algorithms. The need to organise information has motivated the current study to show how Formal Concept Analysis can be used to build a conceptlattice based taxonomy of algorithms. There are many subgraph isomorphism algorithms and it is not feasible to list all of them. As a result, the implications between the attributes of the algorithms are used to build a representative taxonomy of the domain of subgraph isomorphism algorithms. The paper focuses on how the tools from Formal Concept Analysis can be used to build a taxonomy. This is illustrated using a small sample subset of algorithms.

Curricula are not static, especially in Computer Science. Curricula specifications are updated and it is not always clear what the exact updates are nor the impact these updated may have on a curriculum which has been developed to comply with the specification. In this paper, the results of a comparison on the knowledge unit and topic-level between the ACM/IEEE Computer Science curriculum volumes of 2001, 2008 and 2013 is presented.

Many businesses see Big Data and Data Science as a catalyst for innovation. The problem is that many of these businesses are hesitant to embrace these new technologies mainly because of a shortage in skilled manpower. On a global level, higher education institutions are in the process of developing curricula for graduate degree programs relating to Big Data and Data Science. Developing such curriculum has its own challenges. For example: What level of knowledge is required from disciplines such as Computing and Statistics? What underlying foundations in Mathematics are required? This paper presents a framework for the design of an interdisciplinary Big Data and Data Science curriculum on the Master’s level.