Tool wear is the most commonly observed and unavoidable issue in metal milling. The worn or damaged cutting tools will cause materials loss and machines shut-down. To tackle this problem, we propose a new method for predicting the wear condition of end-milling tool. First, we adopt statistic-analysis techniques to analyze the collected data. Second, we select interesting features based on Pearson-Correlation Coefficient (PCC). Finally, those features are applied as inputs to the so called echo-state network to predict subsequent tool wear condition. The experimental results and theoretical analysis both demonstrate that the proposed method performs better than naive Feed-Forward Neural Networks (FFNN) and Time-Series Neural Networks (TSNN).