A Recommender System for Personalized Exploration of Majors, Minors, and Concentrations* Young Park Department of Computer Science & Information Systems Bradley University U.S.A. young@bradley.edu ABSTRACT individual students better find their ideal majors, minors, and concentrations and eventually achieve their academic goals. Many students change their majors during college. Choosing the right Recommender systems provide personalized recommendations academic majors, minors, and concentrations within a major in higher education is challenging but crucial in assuring students’ overall academic of items, such as goods, services, or information, that are most and career success. Personalized prediction of student success in majors, likely to be relevant and interesting to users to help them find the minors, and concentrations will help students better find the right majors, items most useful to them [2,3]. Recommender technologies have minors, and concentrations for them, so as to timely achieve their academic been applied in a variety of application domains including the goals. In this paper, we present a new recommender application for the traditional e-commerce domain as well as emerging domains such academic major/minor/concentration selection problem in the educational as education and engineering. A number of recommender systems domain. The proposed major/minor/concentration recommender system is have been developed and deployed in the educational domain for a collaborative filtering-based recommender based on student grade learning, teaching, and academic advising [4,5]. In the educational prediction, and provides a variety of personalized predictions and recommendations on majors, minors, and concentrations. It is a useful tool setting, the typical users are students, teachers, or academic to guide students, advisors, and administrators during the personalized advisors, and the typical recommendable items are educational major/minor/concentration exploration (declare/change) process. goods, services, or information such as courses, learning materials, topics, student performance, and grades. KEYWORDS In this paper, we describe a new educational recommender Majors, minors, and concentrations; Recommender systems; application in order to help guide the college Collaborative filtering; Personalized grade prediction major/minor/concentration exploration and selection process. The proposed recommender can be used as a tool for students in the major/minor/concentration exploration (declare/change) process 1 INTRODUCTION by providing personalized predictions of success in majors, minors, and concentrations, along with recommendations for individual Selecting the right academic major, minor, and/or concentration students. The proposed tool can also be used by administrators and within a major in higher education is challenging for many advisors to guide decision-making and improve the academic students. However, such selection can be crucial in assuring experience for students at their educational institutions. students’ overall academic success and later career satisfaction and success. Many students change their majors over the course of their college career. Students who pick the right major, minor, and/or 2 PERSONALIZED EXPLORATION OF concentration tend to do better in the requisite courses and thus get MAJORS/MINORS/CONCENTRATIONS better grades. Those students are also less likely to change their chosen major, minor, and/or concentration or drop out, and thus are 2.1 Our Approach more likely to graduate on time. Our goal is to predict how well a student will perform in majors, Students typically choose their major, minor, and/or minors, and/or concentrations, i.e. concentration based on various factors such as their interests, • Given a student and a major, minor, and/or concentration, how will personality, and other considerations [1]. However, we believe that the student perform? it is also very important to consider how students will perform in • Given a student, in what majors, minors, and/or concentrations will that student perform well? their selected major, minor, and/or concentration. Providing An academic major, minor, or concentration typically includes students with personalized prediction of success in majors, minors, course requirements consisting of core courses and elective and concentrations during the exploration process will greatly help courses. A close approximation of a student’s performance can be * Copyright is held by the owner/author(s). RecSys 2017 Poster Proceedings, August 27-31, Como, Italy. ACM RecSys’17, August 2017, Como, Italy Y. Park derived from the grades earned in his or her courses. Thus, we student at that time and (2) compute the grade point average of the represent a student’s performance in a major, minor, or (m+n) courses in the sets in each pair in CMMC × EMMC. concentration as the grades earned in both core and elective courses • Given only a target student S, (1) identify the sets CMMC and EMMC for each potential major, minor, or concentration MMC, (2) predict required in the major, minor, or concentration. Collaborative the personalized grades for courses in the sets in each pair in C MMC filtering recommendation techniques were applied to predict x EMMC that have not taken by the student at that time, and (3) student grades in courses [6,7]. compute the grade point average of the (m+n) courses in the sets in Our approach to major/minor/concentration recommendation is each pair in CMMC × EMMC. based on personalized student-course grade prediction. From our myMMCheck can provide a variety of personalized predictions previous work, we developed a collaborative-filtering-based and recommendations on majors, minors, and concentrations: recommender system called pGPA that predicts personalized • For a student with a specific major, minor, or concentration selected student grades [6]. Given a student, a course and a semester/term, by the student, myMMCheck predicts how well the student will pGPA provides the student’s predicted grade in the course and the perform in that major, minor, or concentration by predicting the best grade prediction proved accurate enough to support high-level overall grade point average along with the list of predicted grades in m core courses and n elective courses. The student will have a better decision making through experimentation on actual student grade perspective on the chosen major, minor, or concentration, and thus data. pGPA uses user-to-user collaborative filtering, item-to-item can better prepare and perform. collaborative filtering, and matrix factorization for personalized • For a student, myMMCheck recommends a list of majors, minors, or course grade predictions. To the best of our knowledge, our concentrations along with the set of m core courses and n elective personalized major/minor/concentration recommender is the first courses in which the student will do well based on the overall grade application of collaborative filtering to the point average predicted. The student can further explore the major/minor/concentration exploration problem. recommended majors, minors, or concentrations. 2.2 A Major/Minor/Concentration 3 CONCLUSION AND FUTURE WORK Recommender We have presented a novel recommender application to the The proposed major/minor/concentration recommender system personalized major/minor/concentration exploration and selection called myMMCheck is a collaborative filtering-based recommender problem. The proposed educational recommender system is based system using the personalized grade prediction system pGPA [6]. on a collaborative filtering-based personalized grade prediction system. The proposed recommender can provide a variety of personalized tips on majors, minors, and concentrations and be useful as a guiding tool during the personalized major/minor/concentration exploration (declare/change) process. As future work, we plan to build a real-world dataset and perform comprehensive testing across curriculums in a higher educational institution. We also plan to incorporate other important factors besides students’ performance, including their interests, personalities, and career goals. REFERENCES [1] Harris-Bowlsbey, J. 2014. White Paper-Career Guidance & Its Implementation in the U.S., Kuder at https://www.kuder.com/research/white- papers/ [2] Ricci, F., Rokach, L., & Shapira, B. 2015. Recommender Systems: Figure 1: The architecture of myMMCheck Introduction and Challenges. In Recommender Systems Handbook (pp. 1-34). Springer US. [3] Felfernig, A., Jeran, M., Ninaus, G., Reinfrank, F., & Reiterer, S. 2013. The overall structure of myMMCheck is depicted in Figure 1. In Toward the next generation of recommender systems: applications and general, a major, minor, or concentration MMC requires m core research challenges. In Multimedia Services in Intelligent Environments (pp. 81-98). Springer International Publishing. courses and n elective courses. Let CMMC be the set of all possible [4] Manouselis, N., Drachsler, H., Verbert, K., & Duval, E. (Ed.). 2013. sets of m core courses {c1, c2, …, cm} and EMMC be the set of all Recommender Systems for Learning. (Springer Briefs in Electrical and possible sets of n elective courses {e1, e2, …, en} required for MMC. Computer Engineering) Berlin/Heidelberg: Springer. [5] Manouselis, N., Drachsler, H., Verbert, K., & Santos, O. C. (Eds.). 2014. Then, the Cartesian product CMMC × EMMC is the set of all possible Recommender Systems for Technology Enhanced Learning: Research Trends pairs of (a set of m core courses, a set of n elective courses) selected and Applications. Springer Science & Business Media [6] Sheehan, M., & Park, Y. 2012. pGPA: a personalized grade prediction tool to from CMMC and EMMC respectively. Each pair consists of (m+n) aid student success. In Proceedings of the sixth ACM conference on courses that meet the requirement of MMC. Recommender systems (pp. 309-310). ACM [7] Elbadrawy, A., Polyzou, A., Ren, Z., Sweeney, M., Karypis, G., & Rangwala, The basic prediction and recommendation process of H. 2016. Predicting Student Performance Using Personalized Analytics. myMMCheck is as follows: Computer, 49(4), 61-69. • Given a target student S and a specific major, minor, or concentration MMC, (1) predict the personalized grades for courses in the sets in each pair in CMMC x EMMC that have not taken by the