Wind loadsWind loads on buildings with non-rectangular plans were investigated through limited experimental and Computational Fluid Dynamics (CFDComputational Fluid Dynamics (CFD)) studies. Therefore, the wind design of non-rectangular buildingsNon-rectangular buildings is described through short guidance in the current building codes and standards. This paper predicts wind loadsWind loads on roofs and walls of non-rectangular buildingsNon-rectangular buildings using Ensemble Machine LearningMachine learning (EML) technique. The EML combines predictions of several regressors, such as Gradient Boosting (GB) and Random Forest (RF), and results in predictions more accurate than the outputs of a single regressor. Numerous tests were performed at the wind tunnelWind tunnel for building models with plan shapes of L, U, T, and X to create a datasetDataset for machine learningMachine learning (ML). An exhaustive grid search with K-fold cross-validationValidation was used for hyperparameters optimization. The Ensemble Machine LearningMachine learning models predicted wind pressure coefficientsPressure coefficients with minimal Mean Squared Error (MSE) and coefficients of determination (R-squared) of up to 0.97.