Forecasting has great importance in the electrical power system. It informs the system in advance about the future load demand and assist the system operator to make decisions regarding spinning reserves, economic dispatch, unit commitment and demand side management etc. Infrastructures of Educational institutions demand power with good quality and reliability. Load demand of Educational institutions differs from other consumers because of sensitive laboratory apparatus, sophisticated instruments and machinery. In order to meet the requirements of Educational institutions, integration of renewable energy sources with national grid require knowledge of accurate demand of energy; a key to building managers and planners. Keeping this in mind we tried to forecast load of an Educational Institute of Pakistan as a case study by exploring decision tree-based algorithm (XGBoost). The results were impressive, and predictions seem near optimal. Previously, this data has never been used in forecasting. However, by training an XGBoost model, short term load forecasting (STLF) was done. The results were determined by calculating MAE, MAPE, R2 and RMSE of predicted and actual values. By efficiently using the renewables during the peak hour, peak load demand can be trimmed.