In order to reduce the sensitivity of the support vector machines ( SVM) to noise and outliers,a new fuzzy least squares-support vector machines classifier based on chaos genetic algorithm is proposed and is abbreviated to FLS-SVMBCGA,in which the clear sets are used to construct a fuzzy membership set and the chaos genetic algorithm is adopted to optimize the parameters. Then,some experiments are carried out on three benchmarking datasets such as the Ripley dataset,the MONK dataset and the PIMA dataset. Finally,the TPD signals from oil and gas transmission pipeline are diagnosed using the proposed classifier. The results show that FLS-SVMBCGA is effective in improving the prediction accuracy of the classification problems with noises or outliers,with a classifying effect for TPD signals being higher than 91.67%,which means that the proposed algorithm can accurately diagnose the TPD signals from oil and gas transmission pipeline.