Shrooq Algarni
Frederick Sheldon
Frederick Sheldon
Submission received: 20 April 2023 / Revised: 21 May 2023 / Accepted: 22 May 2023 / Published: 6 June 2023
Course recommender systems play an increasingly pivotal role in the educational landscape, driving personalization and informed decision-making for students. However, these systems face significant challenges, including managing a large and dynamic decision space and addressing the cold start problem for new students. This article endeavors to provide a comprehensive review and background to fully understand recent research on course recommender systems and their impact on learning. We present a detailed summary of empirical data supporting the use of these systems in educational strategic planning. We examined case studies conducted over the previous six years (2017–2022), with a focus on 35 key studies selected from 1938 academic papers found using the CADIMA tool. This systematic literature review (SLR) assesses various recommender system methodologies used to suggest course selection tracks, aiming to determine the most effective evidence-based approach.
Course recommender systems are considered an application of academic advising systems (AAS). There is currently a need to make sure that students use the information at their disposal to their best advantage in making more informed decisions regarding their academic plans [1]. This is especially important considering the advent of flexible curriculum systems in many educational institutions and the availability of an ever-widening range of courses and programs. In addition to the necessary courses that every student must take, educational institutions often offer a myriad of optional courses (e.g., tech electives). However, most students are unaware of the goals and substance of these courses; thus, they often could have chosen electives better aligned with their academic plans. Another, facet that plays a significant part in this process would be availability which often causes conflicts. Additionally, because students are becoming more numerous and diverse (in terms of their backgrounds, expertise, and ambitions), it is crucial to customize learning and advising procedures more personalized on a student -by-student basis, so that all of these factors can be considered on a more individualized basis tailored to each individuals needs and goals. Moreover, it is certainly doubtful that one learning pathway (i.e., track) would best serve them all [2,3].
Guidance counselors are often employed by educational institutions; they are responsible for assisting students in making their academic decisions. Developmental advising, prescriptive advising, and invasive advising are the three main forms of advising, and each is influenced by the objectives formulated within the advisor–student relationship. This support is focused on the development of an educational partnership between students and their academic advisers in all methods [4]. Advisors help students by assisting them with understanding the university’s educational requirements, assisting them in scheduling the most appropriate modules including prerequisites, introducing them to relevant resources, encouraging leadership and campus involvement, assisting with career development, ensuring that they finish their studies on time, and assisting them in finding ways to make their educational experience personally relevant [5]. However, the counselors are frequently overburdened with too many students and not enough time. Some students can become dissatisfied with the kind of academic guidance the counselors offer. When it comes to comprehending, organizing, and putting ideas for academic achievement into practice, excellent advising produces positive results, whereas poor advising frustrates students and can even be detrimental to their development [6,7] and ultimate success.
A software solution that can manage the advice process effectively and efficiently is needed to assist the educational process and to relieve the educational institutions’ players. A course recommender system can act as a strategic partner in the process of aiding the student (and advisor) in achieving their educational goals and supporting and encouraging their study plan [6,8]. However, unlike most other existing recommendation systems, course recommender systems (CRS) must deal with a sizable decision space that multiplies combinatorically with the number of courses; programs; and the various backgrounds, skills, and goals of a student while simultaneously being subject to numerous restrictions (e.g., maximum credit hour load, course prerequisites, sequencing etc.) [1].
To understand the preferences of various users and forecast products that correspond to their demands, recommender systems scour large databases for important patterns. The word “item” in this context refers to any course, educational component, book, service, application, or product. Machine-learning and data-mining methods are mostly used by CRSs to sort all these constraints toward accomplishing each students’ goals and objectives. Moreover, these same methods are widely utilized in e-commerce and by shops to increase their sales and viewership. These days, those same techniques are being used more frequently for educational recommendation and advising purposes [9], making the whole student-advisor process more effective, efficient and clear cut; win/win for both the institution and its constitutes.
Personalized recommendation systems (PRS) are becoming more and more common in a variety of industries, such as e-commerce, music and video streaming, and they are now making their way into the educational space [3]. These systems strive to make recommendations that are uniquely suited to each user’s tastes and preferences, which makes them extremely pertinent in the context of course recommendation.
The “cold start problem” is a challenge that CRSs encounter. This issue emerges when new students enroll in a program while the CRS lacks sufficient data on them to provide reliable recommendations. Different approaches have been developed to address this issue, which recommender systems often confront in a variety of different industries [10].
A substantial contribution to the field of CRSs is made herein by our analysis. This is accomplished by offering a comprehensive systematic review of the literature (SLR) of empirical research in the field and highlighting the most efficient approaches supported by experiential data. We identify and highlight the knowledge gaps and potential constraints, prompting the community to direct future research endeavors accordingly. Additionally, we explore the difficulties faced by CRSs, particularly those related to handling sizable decision spaces and the cold start problem, providing insightful information about these intricate topics. This study also provides ideas for improving CRSs, considering the growing need for individualized recommendations for fruitfully navigating the complexities of an academic curriculum.
The need for empirical evidence to validate theoretical frameworks that then can be accepted by the scientific community serves as the driving force behind this survey. To the best of our knowledge, no systematic reviews concerning empirical studies in this specific topic exist after a search of the pertinent literature covering the previous six years. As a result, it was necessary to provide the community with an authoritative summary. The purpose of this research therefore is to close or at least lesson that gap.
Any given study’s worth stems from both its inherent qualities and how it complements and advances earlier works. Thus, collecting all the objectives and reliable findings from earlier studies would be a step toward grasping the big picture and forming a roadmap of our consequent knowledge within the field. In a way, the goal of our study was to organize the voluminous quantity of publications by critically examining, assessing, and synthesizing earlier empirical findings.
The absence of empirical validation forms the additional value of research in the field of CRSs in the body of current literature and serves as a valuable result from this study. We combed through the massive body of literature, critically analyzing and evaluating earlier empirical findings to present a comprehensive overview of the existent strategies employed thus far. We draw attention to the positive and negative aspects of earlier research, point out any potential drawbacks, and motivate the research community to rephrase or rethink pertinent study questions and/or hypotheses. The lack of empirical evidence in the existing literature on the added value of research in the specific area of CRSs inspired us to dig deeper. We aimed to sift through the vast body of publications, critically examining, assessing, and synthesizing previous empirical findings to provide an exhaustive account of the applied research approaches used thus far. We highlight the successes and shortcomings of previous studies, identify potential limitations, and seek to inspire the research community to (re)consider these conclusions in designing future studies.
The creation of efficient course recommender systems is necessary due to the expanding complexity of curricula offered and the increased need for individualized learning experiences in educational institutions. While general recommender systems have received a great deal of research, there are not many thorough studies that concentrate solely on course recommender systems. This results from the difficulties this field presents, such as managing prerequisite specifications and a developing course catalog per student growth and shifting educational objectives.
This section will discuss the fundamental questions to be investigated in the study. Two types of research questions must be answered:
