In this work, an accurate method for real time assessment of tool wear in face milling based on the cutting force signals using support vector machine learning is presented. New combinations of signal processing techniques are used to extract relevant signal features, which are robust against random process variations, external chance disturbances and presence of outliers, from the acquired data. Relationship between the extracted features and the observed values of tool wear is built using the support vector regression technique. This machine learning algorithm is accurate and robust for estimation of complex and non-linear relationship among the process variables. Grid search method is adopted to select the optimum model parameters. The developed models are validated on unseen testing data for their predicting capabilities. Significant results are obtained using the proposed method for both laboratory and industrial data.