Algorithms and data structures for automated teaching workload allocation
Date of Issue2017-02-17
School of Computer Science and Engineering
Fair allocation of teaching workload to academic staff is one of the most important administrative duties of an educational institution. However, it is often managed manually with limited data and, often with an arbitrary set of constraints due its complexity, thereby resulting in sub optimal allocations. The objective of this research is to develop efficient methods to automatic the realization of a fair and transparent workload allocation system. A detailed exploration of past attempts, aimed at improving the workload allocation at tertiary institutions, was undertaken to gain a deeper understanding of the various approaches and challenges associated with the process. In particular, it was observed that various approaches were adopted to quantify a variety of components of the workload during the allocation process covering teaching, research and administrative duties. This motivated the development of an efficient time-based workload model called the Workload Unit (WLU) Model as the quantitative framework for measuring and comparing different faculty workloads. This involved the utilization of a combination of weightages corresponding to nature of teaching duties, the number and appointment of faculty as well as their availability for undertaking teaching duties, and attributes (such as class size, level of course, preparation requirements, etc.) of the courses to be offered. In addition, the proposed workload model takes into account the research activeness and administrative duties carried out by each faculty. The resulting workload is separated into two components, namely formal and informal teaching for the purpose of systematic workload allocation. Next, an efficient workload allocation method for formal teaching comprising of lectures, tutorials and labs was proposed by taking into consideration of faculty preference and performance, School policies and priorities and availability of teaching expertise. A novel combined cost function was proposed to determine the feasibility of individual allocations called the Feasibility Index (FI). A greedy approach, with activity-type and course-priority heuristics, is implemented to optimize the quality of the allocations. The proposed approach was validated against the workload allocations made at one of the schools of a University consisting of 100 teaching faculty, 120 courses and 1500 students. The resulting allocations improved workload distribution and quality of assignments over existing manual processes, improving preferences and performance indices by as much as 18% and 7% respectively, while reducing workload variations by 21%. Methods for the automatic translation of the resulting workload allocations into faculty and student centric timetable were proposed. This was achieved using an iterated greedy approach, with an objective to reasonably spread the allocated workload across the week for both faculty as well as students while making efficient use of teaching venues. The proposed approach has been shown to improve the spread of time table allocations across the week by 20% for both students and faculty. Moreover, it has also improved the utilization of the teaching venues by 30%. Realizing Final Year Projects (FYP) allocation is the most challenging part of the informal teaching allocation process, methods for its automation have been proposed by taking into consideration other informal teaching duties such as M.Sc. and Ph.D. supervisions. The proposed methods ensure that the student preference are taken into consideration while adhering to informal workload limits for each faculty. An iterative implementation of the optimal Hungarian algorithm which works on a student-project cost matrix with the preference ranks as the cost of assignment was introduced and multiple iterations were invoked to adjust the cost matrix based on workload constraints to achieve desired results. The resulting solutions comprehensively bettered those achieved by a competing existing system used at a reputable school – improving average student preference for assignments by 10% and workload deviations for faculty by 15%. The above methods were integrated and validated in a real-world scenario. Overview of the technology and platforms used to implement the above solutions have been presented together with screen captures of the platform that is being deployed in the School for past 5 consecutive semesters. This has unreservedly demonstrated the feasibility for automating an inclusive workload allocation process to realize an efficient and fair workload distribution, which is critical to any educational institution. Finally, avenues for future research and improvements are also stated.