Wind turbine design is an iterative process. Many aspects are considered when designing a wind turbine, including aerodynamic and power performance, structural loads and behavior, and control techniques. In the preliminary design stages, the governing equations of each design aspect are used to calculate the different loads and performance outputs while optimizing between them. This is usually made using wind turbine simulation software. This work presents a data-based machine learning (ML) approach towards the design of a micro-scale wind turbine. Extensive simulations are made on a 45 cm diameter rotor while performing parametric analysis using the QBlade wind turbine simulation tool. Different design parameters and wind conditions were changed one at a time, and data were collected to be further analyzed and used to train the ML models. The measurable outputs of the models are the coefficient of power (CP), loads normal and tangential to the blade at midspan (FN and FT), and the torque (T) on the rotor. Linear regression was found unsuitable for predicting CP due to its high nonlinearity; however, it gave satisfactory results for the blade loads. Ensemble models were found to give the highest accuracy for predicting all the desired outputs. The model accuracy is measured in terms of the coefficient of determination (R2), where the model could predict Cp, FN, FT, and T with R2 values of 0.999, 0.984, 0.984, and 0.986 respectively.