DigitalSkills2023:Papers with Abstracts

Abstract. Universities across the globe are experiencing a surge of cyberattacks due to the increased usage of information communication technologies (ICTs). To counteract cyberattacks, universities have implemented cybersecurity measures to ensure that students and the universities’ critical infrastructures are protected. Unfortunately, universities in developing countries continue to face increased cyberattacks despite implementing cybersecurity measures. This study explores the factors that affect students’ compliance with universities’ cybersecurity measures.
The study used a case of the University of Cape Town in South Africa, adopting qualitative research and an interpretive paradigm. We used a deductive approach to theory using Protection Motivation Theory (PMT) as the lens for inquiry. The sample for the study consisted of 40 participants, of which 35 were students and five were staff members of the University. The sample of the study was selected by convenience. We collected empirical data from the participants using semi-structured interviews. The data was then analysed using thematic analysis on NVivo software. The study found that students’ compliance with cybersecurity measures is affected by their perceptions of the seriousness of the threats, the likelihood of the threats happening, their ability to protect themselves against threat, their belief in the effectiveness of the recommended solutions against cyber threats, and the costs associated with compliance to cybersecurity measures. When students perceive the risk as not severe enough to worry about, they do not find it necessary to comply with the University’s cybersecurity measures. Similarly, when the students deem that the recommended compliance actions will not be practical or affordable, they do not adhere to the university cybersecurity measures.

Abstract. The challenge of employee churn is a major issue for most businesses and organizations. Unexpected employee departures can tarnish service delivery, harm customer loyalty, degrade quality of services, drop productivity, and hurt goodwill. The ability to predict employee churn is crucial for retaining valuable employees. This study proposes a predictive model that uses machine learning to forecast employee churn. The predictive model uses feature selection through Pearson correlation methods, information gain, and recursive feature elimination, combined with strong classification methods such as random forest, logistic regression, decision trees, gradient boosting machines, and K- nearest neighbours. The IBM dataset was used for training and testing the proposed predictive model. The accuracy of the different algorithms improved after applying particular feature selection methods. The results yielded showed that the random forest technique outperformed other models in terms of accuracy in the prediction of employee churn.

Abstract. Machine learning (ML) is another branch of technology deemed valuable in the financial sector because of its ability to assist organisations in identifying fraudulent transactions and predicting the ability of customers to repay their bank-issued loans. However, like any type of technology, the adoption of ML introduces changes that impact the processes and operations of the financial service sector. Research on the merits of implementing ML is well captured; however, research on such developments' challenges, issues, and impact is scant. To address this gap, a systematic literature review was undertaken to contribute to the research discourse by investigating the issues, challenges and impacts of implementing ML in the financial business sector. The ScienceDirect, EBSCOhost and ProQuest databases were used to search for the relevant scholarly sources published from 2013-2022. The literature was reviewed based on the PRISMA flow diagram and a thematic analysis of the 35 articles that met the inclusion criteria. The outcome of the review revealed that more complex models, such as artificial neural networks, were implemented in all the identified financial services sectors, followed by support vector machines. This review concludes that the larger the quantity and complexity of financial data, the less the data quality, which significantly reduces the prediction performance, efficiency, and accuracy of the model, which can significantly impact the operations, financial aspects, and the overall reputation of the firms. Future research must explore the impact of ML on the operational, adoption and skills shortages in the financial sector.

Abstract. Global positioning systems (GPS) are more trustworthy than other techniques for gathering location data and have opened new research opportunities. Researchers and partners of Media, Innovation and Communication Technology at Gent University are interested to know how Belgians utilise specific applications on their smartphones. The purpose of this study is to investigate mobileDNA users' behaviour in terms of where, when, and how they utilised their smartphone daily. MobileDNA is a free smartphone logging application that was used as the data collection tool for this research. The GPS metadata is collected every 15 minutes by mobileDNA. We discovered that the application usage sequence of users on a daily path varies from day to day, and that most users tend to travel to more than one city or town in a day. We predicted home of the mobileDNA users and compared smartphone usage at home and outside home and discovered that users spent more time on their smartphone outside their home compared to when they are at home.

Abstract. The COVID-19 pandemic and lockdowns have led to a surge in the use of social media for information sharing and learning about the virus and the vaccine. Our study aimed to understand the sentiments of South African Twitter users towards COVID-19 vaccines. Using the Twitter API version 2, we collected a total of 21,084 tweets from 1 January 2021 to 31 December 2021, during the government's rollout of the vaccine. A sentiment analysis was performed using the VADER lexicon-based classifier to categorise the tweets into positive, neutral, and negative sentiments. The results showed that 40% of the tweets were positive towards the vaccine, 32% were neutral, and 28% were negative. The analysis also revealed that people expressed their opinions on vaccinations more frequently during the early months of the year (January-March 2021) in response to the government's announcement for the vaccine rollout. However, the attitudes towards the vaccine changed throughout the year, indicating that people were sceptical of the government's vaccine rollout strategy, which could have affected the overall vaccine adoption. The findings of this study can provide valuable insights for policymakers and healthcare organisations in shaping effective strategies for promoting vaccine adoption.

Abstract. This paper aims to identify the contributing factors for successful cybersecurity awareness, education, training, programs. The study adopted the systematic literature review method and included 58 primary studies. The study explores approaches for cybersecurity awareness, education and training to improve cybersecurity skills and practice in the extant literature. The study noted several recommendations towards effective cybersecurity awareness, education and training programs. These include considerations focused on the importance of assessing the awareness levels of users, selection of pedagogical approaches, design of the curriculum and supporting organisational and demographic aspects.

Abstract. Data Analytics (BDA) is a crucial component of high-performing e-commerce businesses. In South Africa, where SMMEs are key contributors to economic growth, it is crucial to understand the resources and capabilities they should have in place to adopt BDA and positively impact their business performance. This paper aims to review current literature detailing organisational resources, BDA, and its impact on the business performance of South African e-commerce Small, Medium and Micro Enterprises (SMMEs) and develop a conceptual framework. While various studies show the impact of BDA on business performance in large organisations, and developed and developing economies, the impact it has on business performance, as well as that of organisational resources and entrepreneurial orientation dimensions on South African e-commerce SMMEs, is not well explored. This research adopted a deductive approach, using a systematic approach to literature, 411 journal and conference proceedings articles were retrieved from 2016 to date. 15 articles were selected after analyses and synthesis through a narrative literature approach. Key organisational resources that enable the use of BDA were identified from the literature and used to develop a conceptual framework that can be used for future studies using empirical data. This research is in progress and the preliminary findings were derived through a literature review, that highlights organisational resources such as IT infrastructure, IT human resources, financial resources, risk-taking, innovativeness and proactiveness as relevant to the usage of BDA and its influence on business performance.

Abstract. While data science education may be a solution in democratising data science, there are challenges towards achieving this. One of the challenges is the shortage of qualified academic staff members, who are able to deliver a multidisciplinary curriculum.
Teaching data science is a developing topic and represents a gap in the literature as we embark on the journey of discovering the knowledge required to teach data science. This study aimed to gain insight into instructors’ perceptions on their skills and competencies in teaching data science

Abstract. One of the key goals of higher education institutions (HEIs) is student career success. In HEIs, students are given the necessary subject-specific information, skills, and experience to achieve this. In South Africa, which has historically been one of the technological leaders in Africa, there are currently insufficient Fourth Industrial Revolution (4IR) specialists graduating from higher education to meet the rising need for a skilled workforce in those fields. The completion rate of supplementary online courses to expand this 4IR skills base is also a major concern. In this paper, we report on the strategy implemented to improve the course completion rate of a self-selected sample of students who attended face-to-face Data Science introduction workshops in the Eastern Cape province of South Africa. We achieved a 76% completion rate of two MOOCs courses in these workshops. We examined the contribution of the participant’s personal factors and background contextual factors. We also listed any other factors suggested by the students, which collectively contribute to the learning experience construct of the social cognitive career theory (SCCT) and serve as a practical means of improving 4IR skills exposure and the outcome expectations. Therefore, this paper shows the mechanism which can be used to offer 4IR programmes, in HEIs, to raise students' self-efficacy and outcome expectations in fields such as Data Science. In addition, the learning experience inputs can be utilized to increase students' interest in majoring in 4IR courses.

Abstract. Against the backdrop of escalating contradictions and critiques of the digital economy's trajectory, this study analyzes how the emerging digitalization issues might be philosophically understood from a systems viewpoint. Five systemic digitalization challenges including the circular economy (CE), cyberphysical systems (CPS), sharing economy (SE), digital transformation (DT), and smart systems were identified (SS). To investigate digitalization challenges, the machine, organism, cultural/political, societal/environmental, and interrelationship systems metaphors were used. The machine viewpoint demonstrates that the circular economy challenge may be examined utilizing Hard Systems Thinking (HST) methodologies, with a focus on sufficiency via product design and business model innovation. The organism approach demonstrated how the digital twin notion may be investigated using Socio-Technical Systems (STS) and the Viable System Model (VSM) to diagnose and forecast CPS viability in an increasingly linked Industry 4.5/5.0 environment. In analyzing SE's rentier capitalism, the cultural/political viewpoint demonstrated the applicability of purposeful systems techniques for "people complexity." The societal/environmental viewpoint stressed emancipatory systems approaches to "coercive complexity" as crucial to evaluating the perpetuation of digital exclusion by DT from an emancipatory systems perspective. The interrelationship viewpoint emphasized the significance of systems approaches for researching "structural complexity" in intelligent systems. These viewpoints aid decision- makers in identifying problem-solving strategies based on systems thinking.

Abstract. In poor-resource settings, owning a mobile phone could be an advantage to using developmental interventions based on mobile phones. However, maternal mHealth interventions in these settings are challenged due to low mobile phone ownership among women. Women are less likely to own a mobile phone than their male counterparts. Therefore, for maternal healthcare clients to use maternal mHealth intervention, it is expected that these clients negotiate mobile phone access and usage from owners of mobile phones in their communities. We employed qualitative research methods to understand how maternal healthcare clients who do not own mobile phones negotiate usage of mobile phones for maternal healthcare. Data was collected using semi-structured interviews with maternal healthcare clients and mobile phone owners, and focus group discussions with the maternal healthcare clients. The study found that maternal healthcare clients used cooperative negotiating tactics such as issue-based, compromising, and accommodating to negotiate mobile phone usage. Negotiating mobile phone usage has the potential to enhance digital skills for mobile phone users who do not own mobile phones. The study may inform mHealth implementers on how they may sensitise beneficiaries of mHealth who lack prerequisite technologies on how to negotiate access of mobile phones for mHealth.

Abstract. Heart disease is a major health concern in South Africa, and early and accurate diagnosis is crucial for effective treatment. In this context, dimensionality reduction techniques can play an important role. These techniques can help identify patterns and relationships in large and complex datasets, allowing for more efficient and accurate diagnoses. This paper provides an overview of the use of dimensionality reduction techniques, including principal component analysis (PCA), linear discriminant analysis (LDA), and t distributed stochastic neighbor embedding (t-SNE), in the diagnosis of heart diseases in South Africa. The paper also highlights the importance of considering the interpretability of the results, as well as potential biases in the data and algorithms, when selecting a technique. The purpose of this study is to predict the accuracy of heart disease using Dimensionality technique to determine if there was any enhance in predicting accuracy. Although the SVM show the better accuracy score of 71% over Random Forest with the score of 61% when PCA model is applied the use of dimensional reduction doesn’t produce better results.