This research investigates how effectively machine learning algorithms can predict cutting forces during machining, offering a practical alternative to conventional experimental and numerical methods. The experiments included turning AISI 1117 steel with a cemented car bide insert on a CNC lathe while changing the cutting speed, feed rate, and depth of cut in a planned way. Cutting force data was collected using a Kistler 9257B dynamometer and used to train and test several regression-based machine learning models. These ncluded cubic support vector machine (SVM), gaussian process regression (GPR), various forms of linear regression, decision trees, and ensemble techniques. Two modelling scenarios were analysed: one using cutting speed and feed rate as input variables, and the other using depth of cut as a third input. In the two-variable case, cubic SVM showed the best performance (R²=0.93, root mean squared error [RMSE]=19.57), while GPR with a Matern 5/2 kernel achieved the highest accuracy in the three-variable model (R²=0.99, RMSE<40). Model performance was assessed
using metrics such as R², RMSE, mean squared error, and mean absolute error, with R² and RMSE being most effective for comparisons. The findings indicate that while cubic SVM is suitable for simpler, lower-dimensional data, GPR performs better in capturing complex, nonlinear relationships among multiple variables.