Abstract
<title>Abstract</title> <p> <italic>As one of the leading providers of online education, EduPro had a history of intuitive planning that resulted in several challenges for the company. First, the company appeared to significantly misallocate its marketing efforts due to its undue reliance on data pertaining to past performance rather than current realities. Simultaneously, the company did not utilize faculty expertise optimally, resulting in a significant disparity between the courses on offer and the teachers’ capabilities. The research tackled both issues by designing a predictive analytics model for forecasting at the course level. The project utilized a relational database that stored 10,000 entries arranged in three Excel sheets merged into a 60-row table. The analysis revealed that the allocation of teachers to courses was inconsistent with course revenues, and the Expertise_Match feature introduced to address the discrepancy resolved the issue by matching the teachers’ expertise with the courses’ categories. Lastly, while testing the linear regression algorithm against the more complex Random Forest algorithm, the latter was expected to outperform the first due to utilizing an ensemble learning method, which typically exceeds single-model approaches in performance. However, due to a relatively small sample size (60 rows), Random Forest overfit the training set, performing worse (94%) than the linear regression model (98%). Overall, the dashboard designed and tested in the project is highly beneficial for EduPro as it enables the company to implement the price elasticity function, identify the faculty-course mismatch, and adopt a quantitative approach to forecasting that adheres to the literature.</italic> </p>