Article Preview
Top1. Introduction
In a fast-changing global environment, the adoption of higher education to the challenges in the internal as well as the external environment requires a multi-layer approach for the enhancement of strategic decision-making. A variety of educational data are available for processing and for supporting decision making. The availability of structured and unstructured data related to different stakeholders in the academic context set significant challenges and opportunities for their use for enhanced decision making. While in the past educational, institutional, and academic data were stored in diverse, stand-alone systems, nowadays the integrative decision-making process requires integrative data warehouses and advanced learning techniques. A new research domain, called learning analytics has emerges in the last years, providing advanced, systematic research for the use of education data towards improvement in the efficiency of higher education institutions. Several structured, un-structured and semi-structured decisions in higher education context, pose significant challenges for the analytical processing of relevant data. The aggregation of data, their availability for decision making processes requires well defined computational methods for the codification, structuring and analysis of data. The evolution of data-warehouse technology in academia in the last years incorporates flexible analytical tools towards enhanced descriptive and predictive processing of educational and academic data. The latest developments in the domains of artificial intelligence and business analytics in the last years promote a new era of decision-making (Lytras et al., 2017). The key characteristics of this evolution in the domain of higher education can be characterized as follows:
- •
The decision-making process becomes more complicated. A number of structured and unstructured data provide the basis for advanced students profiling towards personalized services
- •
The descriptive and predictive data analytics methods offer unforeseen opportunities for the execution of models. Within this context the exploitation of semi-structured decision-making capabilities requires a balanced risk approach. The predictability mechanism of models can be enhanced by AI algorithms in order to provide an ongoing evolutionary better performance
- •
The decision-making eco-system in higher education requires new visualization approaches and sophisticated dashboards
- •
From a computational point of view the development of advanced artificial intelligence approaches requires a holistic approach to the data management of critical bid data within and outside of the academic institutions.
- •
The portability of data as well as the portability of any advanced decision method poses critical requirements for cloud-based solution.
Within this extremely challenging research environment for the application of predictive analytics in the context of higher education our approach is promoting the idea of a decision maker, which integrates academic metrics towards prediction of academic performance. In the basis of our approach a set of structured academic data are exploited in order to inform a predictive model. This model supports an integrated ecosystem of academic decisions.
This research work is organized as follows: In the next section we elaborate on the recent literature and the state of the art related to learning analytics as well as to the adoption of information processing over educational data for decision making. Then in section 3, we analyze the research methodology and the research objectives of our study. In section 4, we present the empirical testing and the analysis of our neural-network approach for academic performance forecasting. Then in the next section we provide a discussion for the implications of our approach and our contribution to the body of the knowledge in the domain. Limitations of the study and future research directions are finally summarized.