M.Parvathi, Head, Dept of Computer Applications, Senthamarai College of Arts and Science, Madurai, India 625016

Dr. S. Thabasu Kannan, Principal, Pannai College of Engineering and Technology, Sivagangai, India 63056

Recent technological advances have tremendously increased the amount of collected data. Besides the total amount of collected information, the complexity of data objects increases as well. With the growing amount of data the complexity of data objects increases. Modern methods of KDD should therefore examine more complex objects than simple feature vectors to solve real-world KDD applications adequately. To analyze these data collections, new data mining methods are needed that are capable of drawing maximum advantage out of the richer object representations. This paper contributes to the field of data mining of complex objects by introducing methods for clustering and classification of compound objects. The area of KDD deals with analyzing large data collections to extract interesting, potentially useful, so far unknown and statistically correct patterns. The data objects to be analyzed are complex. Thus, they are not represented in the best possible way by the common approach using feature vectors. So, data mining algorithms have to handle more complex input representations. KDD is necessary to analyze the steadily growing amount of data caused by the enhanced performance of modern computer systems. Multi-represented objects are constructed as a tuple of feature representations where each feature representation belongs to a different feature space. The goal of clustering multi-represented objects is to find a meaningful global clustering for data objects that might have representations in multiple data spaces. This paper contributes to the development of clustering and classification algorithms that employ more complex input representations to achieve enhanced results. Therefore, the paper introduces solutions for real-world applications that are based on multi-represented object representations. To analyze multi-represented objects, a clustering method for multi-represented objects is introduced that is based on the projected-based clustering algorithm. This method uses all representations that are provided to find a global clustering of the given data objects. To map new objects into ontology a new method for the hierarchical classification of multi-represented objects is described. The system employs the hierarchical structure of the efficient classification method using support vector machines.