In order to judge the development trend of the state of optical fiberScomposite submarine cable and send timely the fault warning signals, the multiSstep forecast modelSof the distributed fiber-optic data based on weighted least squares support vector machine(WLS-SVM) is proposed. HierarchicalSthreshold method is used to calculateSthe threshold on different scales which is derived from wavelet decomposition of the measured data. SoftSthreshold method can be used to eliminate the influence ofSrandomSnoise onSthe accuracy of prediction. The phase space is reconstructed using G-P algorithm based on chaotic sequence analysis, then the optimalSembedding dimension can be determined and the chaotic characteristicsSof time seriesSof frequencySshift can be verified.STheSphase pointSof reconstructed phase spaceSis fed to theSWLS-SVM model to proceed recursiveSmulti-step prediction. Finally, 6 step prediction of frequency shift in two typical position of cable is carried out. The resultsSshow thatSthe maximum average relative error of recursiveS6Sstep predictionSis 1.80%,Swhich it hasShigher prediction accuracySand better applicability compared with theSstandardSsupport vector machine method.