Data-Driven Tutor Scheduling at WARC: Time-Series Modeling and Predictive Analytics for Academic Support Optimization

Abstract

This study aims to assist the Writing and Academic Resources Center (WARC) at our university in developing a data-driven approach for allocating tutor schedules across subjects. WARC currently distributes tutor hours without a clear method for estimating the percentage of time each subject should occupy in a semester. To address this gap, we analyzed historical tutoring data from 2020 to 2025, including planned sessions and actual booked sessions. Forecasts for future demand were generated using the Prophet model with an added Spring-semester regressor. In addition to semester-level forecasting, we conducted an intra-semester analysis to examine weekly booking dynamics and identify peak load periods. To complement the quantitative findings, we also administered a Google Forms questionnaire to current tutors. The survey collected tutors’ perceptions of booking patterns, seasonal demand differences (Fall vs. Spring), peak weeks and hours, perceived difficulty of subjects, and reasons for low or high demand. The final output includes visualizations, insights on weekly and semester-level demand, and an Excel table recommending the percentage share each subject should occupy for the next four semesters, along with expected increases or decreases in overall bookings. This work provides WARC with clear, evidence-based guidance to optimize tutor scheduling and better prepare for fluctuations in student demand