The prediction of transmission quality(QoT) is becoming more and more important in optical networks. Machine learning has become an important means to realize the prediction of QoT in optical networks in the future. This paper presents a new technology of QoT prediction based on machine learning classifier. Through the data generated by the transfer equation, it can be used for classifier training and performance testing. The performance of K-Nearest Neighbor(KNN), logistic regression(LR) and support vector machine(SVM) are verified by simulation. The simulation results show that, compared with the traditional QoT estimation method, the machine learning based on method can effectively reduce the computational complexity and provide quite high prediction accuracy. This method is a new QoT estimation scheme with broad application prospects.