Data-driven methods along with a machine learning approach play an important role in analyzing data to extract scientifically interesting patterns between tribological and material properties due to the varying composition of the friction material (FM). The effects of tribological test variables (speed, and deceleration), and material properties (density, water porosity, oil porosity, thermal conductivity, compressibility, hardness, and acetone extraction) on the friction of the friction material due to varying composition of the ingredients were investigated using data-driven methods. For the study using a data-driven approach, the experimental results are collected for the conventional brake for various friction materials tested over the brake inertia dynamometer. To predict the COF of the friction material, four different machine learning (ML) algorithms, including K Nearest Neighbors (KNN), Support Vector Machine (SVM), Random Forest (RF), and Gradient Boosting Machine (GBM), were applied to tribological experimental data of the friction material of various compositions. We showed through performance analysis that the ML models can accurately predict friction of friction material from data on the material and tribological test variables. It was clear from a comparison of model results that the Gradient Boosting regressor (GBR) performed better at COF prediction than other ML models. According to a further investigation of feature importance, acetone extraction, and deceleration have the most impact on forecasting the COF of composites.