Article Preview
TopIntroduction
The technological development of recent times has facilitated access to information. People started to look for new products, services and new forms of knowledge. Online Education takes advantage of this new proactive paradigm of search for knowledge, providing learning materials, which comprise complete courses.
For many years, the technologies based on the improvement of education had as main objective the transfer of information from the central figure of a teacher whose main activity is to provide learning materials to be passively consumed by the students. Most of the courses offered on traditional Virtual Learning Environments (VLE) and Massive Open Online Courses (MOOCs) are still based on the passive form of content delivery. Thus, a learning concept is taught in the same way, with the same learning materials for all students (da Silva Lopes & Fernandes, 2009; Acampora et al., 2011). Therefore, relevant omissions, such as the student learning experience has been considered one of the main obstacles to the adoption of these systems (Acampora et al., 2011; Dwivedi et al., 2018).
Adaptive Learning (AL) is a set of techniques aimed at providing students with a personal and unique experience seeking to maximize their learning performance (Phobun & Vincheanpanya, 2010). Therefore, it may be considered to addresses some of the aforementioned issues. One of the most researched field in AL is the learning material recommendation (Dwivedi & Bharadwaj, 2013; Erdt et al., 2015). However, the recommendation by itself does not guarantee overall understanding (Dwivedi et al., 2018), since it does not consider a sequential organization in which each learning material must be in agreement with the previous and the next one (Durand et al., 2013). Therefore, curriculum sequencing – or Adaptive Curriculum Sequencing (ACS) – is a crucial issue in which the purpose is to help the system to find the best sequence of learning materials that meet the student profile (Premlatha & Geetha, 2015). The challenge lies in the construction of this sequence, since unsuitable sequences can lead to an increase in failure and dropout rates in the course (Xie et al., 2017).
Learning materials repositories, used by e-learning systems, are getting larger due to the ease of creation and access to digital content and the Open Educational Resources (OERs) movement. Moreover, several parameters are involved in the adaptive process, making the decision process more difficult. Thus, finding an optimal curriculum sequence is a combinatorial problem falling in the NP-Hard class of problems (Pushpa, 2012). It is worth noting that this fact motivated several authors to use heuristics and metaheuristics, especially of Evolutionary Computation approaches, to address the ACS problem, since in a classical manner, they are alternatives to similar problems (Al-Muhaideb & Menai, 2011; Wang & Wu, 2011, Pushpa, 2012; Khamparia & Pandey, 2015). New evolutionary approaches have been introduced in recent years with promising results in similar problems. Exploring such new approaches can bring benefits in solving ACS problem.
The process of adapting at the curriculum sequencing level relies heavily on the fitness criteria. Recently, several papers have explored solutions using a variety of parameters (Al-Muhaideb & Menai, 2011; Wang & Wu, 2011; Pushpa, 2012; Khamparia & Pandey, 2015). These works have shown that most of the learning materials parameters, such as the difficulty and the average time of learning are defined through consensus of experts. On the other hand, modeling student's profile is more complex and is usually based on pretests or even on logs associated with the student's past behavior in the system. There is also the complexity of associating student's model with knowledge domain, for instance, by trying to associate the student's Learning Style (LS) with learning material features. Due to these difficulties, it is noticed that the intrinsic parameters related to students are not used in the majority of the works.