Mining Multivariate Temporal Patterns for Event Characterization and Prediction

Characteristic and prediction of the events are essential in many applications, such as forecasting economic growth, financial decision making etc. This can be done by processing the temporal patterns which are observed event data sequence often closely related to certain time-ordered structures. Among several existing method reconstructed phase space work well but only for univariate data sequence. So we propose a multivariate reconstructed phase space which is uses supervised clustering for characteristic and prediction of event from these dynamic data sequence. An optimization method is applied finally to estimate the parameters of the classifier that defines an optimal decision boundary in the