Abstract—In Peru, it has been observed that the academic performance of school students is concerning, with many cases falling below the average of OECD countries. These figures become even more alarming upon closer inspection: about 66% of Peruvian students perform poorly in math, and approximately 50% face similar challenges in reading. Several studies have applied forecasting techniques in the educational field aimed at improving academic performance, ranging from artificial neural networks to machine learning models focused on linear regression. The Holt-Winters Additive model has also proven to be a highly accurate forecasting tool. In the present research, two tests (Math forecast score and Language forecast score) were conducted to predict grades in the 6th grade section of a private school. Additionally, a residual analysis was performed to validate the robustness of the forecasting model. The forecast results showed experimental errors of 3.60% for Math and 4.65% for Language. In both cases, the residuals were very robust, providing reliability to the applied model. The application of this statistical model in the field of education can be considered to optimally manage resources in advance, thereby addressing academic performance issues in schoolchildren. 

Keywords—academic performance, school students, Holt-Winters Additive, forecast, residual analysis