Extending Support Vector Machines to Discover Temporal Periodic Patterns

We introduce an extension of the support vector machine (SVM) method to discover temporal periodic patterns. This extension, v-SVCM, uses a parameter v to characterize confidence and accuracy of pattern discovery. We apply the v-SVCM method empirically to the stock price data of two Chinese companies. The results show that the value of the parameter v is a decisive factor in determining the confidence degree in the process of temporal periodic pattern discovery. The results also show the important role played by the choices of discretisation intervals and classification standards in the discovery of temporal periodic patterns.