Analysis and modeling of sequential pattern as multimedia data representation

The goal of feature extraction in multimedia mining is to discover important features for represented into a form that can represent information of multimedia data. Sequential pattern is one form of data representation formed of a number of elements that appear in sequence. The goal of this study is to analyze sequential pattern representation performancy to improve accuracy and efficiency. Analysis was performed by comparing performance of WordNet and sequential patterns representation from text documents. And comparing performance of frequent local histograms and sequences of feature sets representation for image. Performance of representation is semantically meaningful information from each representation, and the number of features that are formed and the process of forming a representation. Analysis was also conducted on performance of sequential patterns as multimedia data representation, to see what characteristics are influential in improving the accuracy and efficiency. Beside analyzing the concept, also analyzed mathematically by using formal concept analysis. Modeling done on a sequential pattern characteristics that affect the maintenance of semantic meaning and increase time efficiency through several stages of modeling : definition of context, formal concept formation, and visualization through the concept lattice. Based on analysis and modeling that has been done, showed that the performance of sequential patterns as multimedia representation is better than the other representation, in keeping the semantic meaning of document and increase efficiency. Sequential patterns characteristics that influence the improving quality is obtained through its ability to generate sequential information, temporal information, and sequential information that have gaps.